Advanced Morphometric Techniques Applied to The

UNIVERSIDAD POLITÉCNICA DE MADRID ESCUELA TÉCNICA SUPERIOR DE INGENIEROS DE TELECOMUNICACIÓN

ADVANCED MORPHOMETRIC TECHNIQUES APPLIED

TO THE STUDY OF HUMAN BRAIN ANATOMY

TESIS DOCTORAL

Yasser Alemán Gómez Ingeniero en Tecnologías Nucleares y Energéticas Máster en Neurociencias

Madrid, 2015

DEPARTAMENTO DE INGENIERÍA ELECTRÓNICA ESCUELA TÉCNICA SUPERIOR DE INGENIEROS DE TELECOMUNICACIÓN

PHD THESIS

ADVANCED MORPHOMETRIC TECHNIQUES APPLIED

TO THE STUDY OF HUMAN BRAIN ANATOMY

AUTHOR Yasser Alemán Gómez Ing. en Tecnologías Nucleares y Energéticas MSc en Neurociencias

ADVISOR Manuel Desco Menéndez, MScE, MD, PhD

Madrid, 2015

Departamento de Ingeniería Electrónica Escuela Técnica Superior de Ingenieros de Telecomunicación Universidad Politécnica de Madrid

Ph.D. Thesis Advanced morphometric techniques applied to the study of human brain anatomy Tesis doctoral Técnicas avanzadas de morfometría aplicadas al estudio de la anatomía cerebral humana

Author: Yasser Alemán Gómez Advisor: Manuel Desco Menéndez

Committee:

Andrés Santos Lleó Universidad Politécnica de Madrid, Madrid, Spain Javier Pascau Gonzalez-Garzón Universidad Carlos III de Madrid, Madrid, Spain Raymond Salvador Civil FIDMAG – Germanes Hospitalàries, Barcelona, Spain Pablo Campo Martínez-Lage Universidad Autónoma de Madrid, Madrid, Spain Juan Domingo Gispert López Universidad Pompeu Fabra, Barcelona, Spain María Jesús Ledesma Carbayo Universidad Politécnica de Madrid, Madrid, Spain Juan José Vaquero López Universidad Carlos III de Madrid, Madrid, Spain

Esta Tesis ha sido desarrollada en el Laboratorio de Imagen Médica de la Unidad de Medicina y Cirugía Experimental del Instituto de Investigación Sanitaria Gregorio Marañón y en colaboración con el Servicio de Psiquiatría del Niño y del Adolescente del Departamento de Psiquiatría del Hospital General Universitario Gregorio Marañón de Madrid, España.

Tribunal nombrado por el Sr. Rector Magnífico de la Universidad Politécnica de Madrid, el día ___ de ______de 2015.

Presidente: ______

Vocal: ______

Secretario: ______

Suplente: ______

Realizado el acto de defensa y lectura de la Tesis el día ___ de ______de 2015, en la E.T.S.I de Telecomunicación.

Calificación: ______

______

EL PRESIDENTE LOS VOCALES

______

EL SECRETARIO

“Si saber no es un derecho, seguro será un izquierdo” Silvio Rodríguez (El escaramujo)

Agradecimientos

A unas horas de que este documento se entregue definitivamente lo más simple hubiera sido no escribir agradecimientos y así evitar la posible injusticia de no mencionar a todos los que me han ayudado, de una manera u otra, a llegar hasta aquí. Precisamente, después de este esfuerzo, es evidentemente que podemos sobrevivir unas horas más sin optar por lo simple. Agradecer ante todo a Manolo, primero por la oportunidad que me dio al llegar a España de formar parte de este grupo y segundo por la ayuda que me ha brindado durante estos años en el laboratorio. Agradecer la ayuda para terminar este trabajo y también, y a lo cual doy mayor importancia, por lo que he aprendido acerca del esfuerzo que se debe realizar para mantener a flote un laboratorio del cual, todos los que estuvieron, estamos o estarán, tendremos siempre un recuerdo muy especial. Es una tarea casi quijotesca mantener un grupo así con lo poco valorada que está la investigación actualmente. Además quisiera agradecer a otras dos personas muy importantes para llevar este trabajo a algún lugar. Primeramente a mi “nederlandse vriend” Joost. Sin ti nada de lo que se leerá de aquí en adelante hubiera existido. Gracias por esas discusiones, críticas directas y evidentemente por la amistad que mantenemos. En segundo lugar a Chus, sin la cual hubiera estado perdido durante todo este proceso, especialmente al principio. En el medio y al final también lo hubiera estado, seamos sinceros. Un agradecimiento muy especial a mis padres que siempre me han apoyado y de los cuales cada día me siento más orgulloso. De ellos he absorbido todo lo que he podido para llegar a ser mejor persona. De mi madre he heredado la cordialidad, la facilidad para hacer amigos y el sacrificio, de mi padre el despiste, la despreocupación material y el soltar lo que se piensa. A mis primos y a su camioneta. Mi prima, fruto de la unión de dos personas espectaculares, es, resumiendo, mi hermana mayor. Mi primo, antes que su esposo, es, si no la más noble, una de las personas más nobles de mi planeta. Gracias a ambos por todo lo que han hecho por nosotros y por darme el hermano-primo pequeño (ya sólo de edad) que tengo. Estoy super orgulloso de haber suspendido el examen final de mecánica teórica en 2do año de carrera y espero que el “reggaetón” sea como el asma, que se pasa con la edad.

Agradecer también a mis primos “italianos” Belkis, Franco e Illide y a la familia Osvaldos-Ariel-Kenia-Teida de 9na entre F y G. El hecho de ser del Vedado influyó siempre en la baja calidad de juego en la pelota a la mano pero siempre estuve ahí para levantar el nivel. El Cerro es otro “level”. A mi chica por irse conmigo aunque finalmente he terminado viniendo yo. Gracias por la confianza, el apoyo y el amor que me has dado durante todo este tiempo. El venir me ha dado la posibilidad de conocer a tu familia que también, como no podía ser de otra manera, es espectacular. Incluyo a Santi dentro de tu familia porque evidentemente forma parte de ella. Una de las ventajas que tiene emigrar es que conoces mucha gente y muchas de ellas se convierten luego en nuevos amigos. Ahora tengo el doble que hace 7 años. Agradecer a todos ellos el estar siempre ahí y que sepan que yo también estoy aquí para cualquier combate. Agradecer a Velarde, a Erick, al Yaci, a Ramón, al Dani, a los Alejandros (Alejandro Ojeda y Alejandro Fiel), Betty, Rogney, Joost, Sonia, Monzón, Ricardo, Iván Linares, Maricela, Maylen, Yanays, Diamela, Jose Luis, Sandra, Pedrito, Eduardo, Lester, Nelson, Tin, Joelay, Mónica, Ana, Pablo, Javi, Cris, Alex/Sisni, Trajana, Esther, Oscar, Lorena, Fidel, Susana, David, Anabea, Kenia, Kike, Vero, Paula, Inma, Carmen, Iván, Eva, Javi Pascau, Juanjo, Pedro, Evan, María, Chema, Eu, etc, etc, etc … Quisiera finalmente agradecer la suerte de haber podido conocer a una persona increíble. Gracias por todo Santi Reig, nos hiciste mejores. Un honor haberte conocido. ><((((º>`·.¸¸.·´¯`·.¸.·´¯`·...¸><((((º>`·.¸¸.·´¯`·.¸.·´¯`·...¸><((((º> `·.¸¸.·´¯`·.¸.·´¯`·...¸><((((º>`·.¸¸.·´¯`·.¸.·´¯`·...¸><((((º>`·.¸¸.·´

Contents

Contents Abstract ...... v Resumen ...... vii List of tables...... ix List of figures ...... xi Chapter 1: Motivation and objectives ...... 1 1.1. Motivation ...... 1 1.2. Hypothesis and objectives ...... 3 1.3. Outline ...... 3 Related references ...... 5 Chapter 2: State of the art in brain morphometry ...... 7 Introduction ...... 7 2.1. Preprocessing steps in brain morphometry ...... 7 2.1.1. Inhomogeneity correction ...... 7 2.1.2. Image registration and spatial normalization ...... 8 2.1.3. Tissue classification ...... 10 2.1.4. Structure parcellation ...... 11 2.1.5. Cortical surface extraction ...... 12 2.1.6. Volumetric and geodesic smoothing ...... 14 2.2. Region of interest analysis...... 15 2.3. Voxel based morphometry ...... 15 2.4. Surface based morphometry or vertex-wise analysis ...... 17 2.5. Structural shape analysis ...... 19 2.6. Sulcal morphometry ...... 21 Related references ...... 23 Chapter 3: Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis ...... 27 Abstract ...... 29 3.1. Introduction ...... 30 3.2. Materials and methods ...... 31 3.2.1. Participants ...... 31 3.2.2. Clinical assessment ...... 32 3.2.3. Medication ...... 33 3.2.4. MRI Acquisition ...... 34 3.2.5. Image Analysis ...... 35 3.2.6. Statistics ...... 38

i Contents

3.3. Results ...... 39 3.3.1. Demographic and clinical data ...... 39 3.3.2. Whole thalamic volume ...... 40 3.3.3. Regional thalamic volume ...... 40 3.3.4. Correlation with clinical variables ...... 42 3.4. Discussion ...... 42 Acknowledgements ...... 45 Related appendices ...... 46 Appendix 3.A. Thalamic volume assessment and results using FSL-FIRST ...... 46 Appendix 3.B. Determination of an adequate combination of parameters for generating spherical harmonics (SPHARM)-based thalamic surfaces ...... 48 Related references ...... 50 Chapter 4: The human cortex flattens during adolescence ...... 53 Abstract ...... 55 4.1. Introduction ...... 56 4.2. Materials and methods ...... 57 4.2.1. Clinical and functional assessment ...... 58 4.2.2. MRI Acquisition ...... 59 4.2.3. Image analysis ...... 60 4.3.4. Statistics ...... 64 4.3. Results ...... 65 4.3.1. Longitudinal changes in cortical thickness (CT), surface area (SA), gyrification index (GI), hull surface area (HS), white matter thickness (WT), sulcal depth (SD), sulcal length (SL), and sulcal width (SW) ...... 65 4.3.2. Differences in longitudinal changes between lobes...... 65 4.3.3. Partial correlations between longitudinal changes with each lobe ...... 66 4.4 Discussion ...... 67 Acknowledgements ...... 70 Related appendices ...... 71 Appendix 4.A. Manual labeling of sulci generated by the BrainVISA software into lobes . 71 Appendix 4.B. Studying and removing the scanner effect over cortical metrics ...... 73 Related references ...... 76 Chapter 5: Cortical morphology of adolescents with bipolar disorder and with schizophrenia ...... 79 Abstract ...... 81 5.1. Introduction ...... 82

ii Contents

5.2 Materials and methods ...... 83 5.2.1. Subjects sample ...... 83 5.2.2. Clinical, functional and cognitive assessment ...... 84 5.2.3. MRI acquisition ...... 85 5.2.4. Image analysis ...... 86 5.2.5. Statistics ...... 86 5.3. Results ...... 89 5.3.1. Demographic and clinical differences ...... 89 5.3.2. Group differences in cortical thickness, surface area, gyrification index, and sulcal width ...... 92 5.3.3. Correlation between frontal gyrification index and sulcal width in the patient group ...... 94 5.3.4. Correlations between the morphometry measurements and clinical variables ...... 94 5.4. Discussion ...... 94 Related appendices ...... 98 Appendix 5.A. Quality control and final sample ...... 98 Related references ...... 99 Chapter 6: Conclusions and future work ...... 103 6.1. Original contributions and general discussion of this thesis ...... 103 6.2. Conclusions ...... 106 6.3. Future research lines ...... 107 Related appendices ...... 108 Appendix 6.A. Image processing platform ...... 108 Related references ...... 112 List of publications ...... 113

iii

Abstract

The ever evolving sophistication of magnetic resonance image techniques continue to provide new tools to characterize and quantify, in vivo, brain morphologic changes related to neurodevelopment, senescence, learning or disease. The majority of morphometric methods extract shape or size descriptors such as volume, surface area, and cortical thickness from the MRI image. These morphological measurements are commonly entered in statistical analytic approaches for testing between-group differences or for correlations between the morphological measurement and other variables such as age, sex, or disease severity. A wide variety of morphological biomarkers are reported in the literature. Despite this wide range of potentially useful biomarkers and available morphometric methods, the hypotheses and findings of the grand majority of morphological studies are biased because reports assess only one morphometric feature and usually use only one image processing method. Throughout this dissertation biomarkers and image processing strategies are combined to provide innovative and useful morphometric tools for examining brain changes during neurodevelopment. Specifically, a shape analysis technique allowing for a fine-grained assessment of regional thalamic volume in early-onset psychosis patients and healthy comparison subjects is implemented. Results show that disease- related reductions in global thalamic volume, as previously described by other authors, could be particularly driven by a deficit in the anterior-mediodorsal and pulvinar thalamic regions in patients relative to healthy subjects. Furthermore, in healthy adolescents different cortical features are extracted and combined and their interdependency is assessed over time. This study attempts to extend current knowledge of normal brain development, specifically the largely unexplored relationship between changes of distinct cortical morphological measurements during adolescence. This study demonstrates that cortical flattening, present during adolescence, is produced by a combination of age-related increase in sulcal width and decrease in sulcal depth. Finally, this methodology is applied to a cross-sectional study, investigating the mechanisms underlying the decrease in cortical thickness and gyrification observed in psychotic patients with a disease onset during adolescence.

Resumen

El desarrollo de las técnicas de imágenes por resonancia magnética han permitido el estudio y cuantificación, in vivo, de los cambios que ocurren en la morfología cerebral ligados a procesos tales como el neurodesarrollo, el envejecimiento, el aprendizaje o la enfermedad. Un gran número de métodos de morfometría han sido desarrollados con el fin de extraer la información contenida en estas imágenes y traducirla en indicadores de forma o tamaño, tales como el volumen o el grosor cortical; marcadores que son posteriormente empleados para encontrar diferencias estadísticas entre poblaciones de sujetos o realizar correlaciones entre la morfología cerebral y, por ejemplo, la edad o la severidad de determinada enfermedad. A pesar de la amplia variedad de biomarcadores y metodologías de morfometría, muchos estudios sesgan sus hipótesis, y con ello los resultados experimentales, al empleo de un número reducido de biomarcadores o a al uso de una única metodología de procesamiento. Con el presente trabajo se pretende demostrar la importancia del empleo de diversos métodos de morfometría para lograr una mejor caracterización del proceso que se desea estudiar. En el mismo se emplea el análisis de forma para detectar diferencias, tanto globales como locales, en la morfología del tálamo entre pacientes adolescentes con episodios tempranos de psicosis y adolescentes sanos. Los resultados obtenidos demuestran que la diferencia de volumen talámico entre ambas poblaciones de sujetos, previamente descrita en la literatura, se debe a una reducción del volumen de la región anterior-mediodorsal y del núcleo pulvinar del tálamo de los pacientes respecto a los sujetos sanos. Además, se describe el desarrollo de un estudio longitudinal, en sujetos sanos, que emplea simultáneamente distintos biomarcadores para la caracterización y cuantificación de los cambios que ocurren en la morfología de la corteza cerebral durante la adolescencia. A través de este estudio se revela que el proceso de “alisado” que experimenta la corteza cerebral durante la adolescencia es consecuencia de una disminución de la profundidad, ligada a un incremento en el ancho, de los surcos corticales. Finalmente, esta metodología es aplicada, en un diseño transversal, para el estudio de las causas que provocan el decrecimiento tanto del grosor cortical como del índice de girificación en adolescentes con episodios tempranos de psicosis.

vii

List of tables

Chapter 3:

Table 3.1: Demographic and clinical variables of the adolescent males with early-onset psychosis and the adolescent male controls...... 34

Table 3.2: Intracranial volume and total left and right thalamus volume of the adolescent males with early-onset psychosis and controls...... 39

Chapter 4:

Table 4.1: Sociodemographics at baseline of the sample...... 58

Table 4.2: Acquisition parameters as well as number of subjects for each scanner...... 59

Table 4.3: Percentage change relative to baseline per lobe over a two-year period in 52 healthy adolescents (age range 11-17 years)...... 65

Chapter 5:

Table 5.1: Sociodemographic and clinical characteristics at baseline of healthy controls, early-onset psychosis (EOP) patients with a two-year follow-up diagnosis of schizophrenia and EOP patients with a 2-year follow-up diagnosis of bipolar disorder with psychotic symptoms...... 91

Table 5.2: Pair-wise comparisons between healthy controls (n=52), early-onset psychosis (EOP) patients with a two-year follow-up diagnosis of schizophrenia (n=20) and EOP patients with a 2-year follow-up diagnosis of bipolar disorder with psychotic symptoms (n=20)...... 93

Table 5.A.1: Sociodemographic and clinical characteristics at baseline of included and excluded healthy controls and early-onset psychosis (EOP) patients...... 98

List of figures

Chapter 2:

Figure 2.1: Inhomogeneity artifact in a T1-weighted MRI...... 8

Figure 2.2: Linear (affine) transformations. A) Translation. B) Rotation. C) Scaling. D) Shearing. E) Perspective deformation...... 9

Figure 2.3: Nonlinear registration. A) T1-weighted MRIs from two subjects. B) Image overlap after an affine registration. As can be observed the corpus callosum (in red) from both MRIs does not coincide. C) Zoom of the region below the red rectangle. The green and blue dots are corpus callosum landmarks in both subjects where their spatial coincidence was not achieved by the affine registration. D) Perfect coincidence of the corpus callosum from both subjects after the nonlinear registration process. The yellow vectors show the distance that each green point needs to cross to reach its corresponding blue point...... 9

Figure 2.4: Brain tissues segmentation from T1-weighted images. Red: Cerebrospinal fluid. Magenta: White matter. Green: Grey matter...... 10

Figure 2.5: Flowchart showing the steps followed by registration based algorithm for the brain structures parcellation...... 11

Figure 2.6: Pial and white surfaces extracted from T1-weighted MRI for both hemispheres. Note: Pial surface is the GM/CSF boundary and white surface is the WM/GM boundary. 12

Figure 2.7: Hemispheric pial and hull surfaces. The regional gyrification index is the quotient between the area of the pial and hull surfaces...... 13

Figure 2.8: Pial, inflated and sphere surfaces overlaid with cortical parcellation (A), thickness (B), curvature (C) and local gyrification maps (D). Note: Inflated surface is used to show the areas buried inside sulci...... 13

Figure 2.9: Smoothing. A) Volumetric smoothing. B) Geodesic smoothing...... 14

Figure 2.10: Voxel based morphometry flowchart...... 16

Figure 2.11: Possible causes of the differences obtained during a VBM analysis...... 17

Figure 2.12: Volume as a composite measure of surface area and cortical thickness...... 17

Figure 2.13: Surface alignment and normalization strategy for cortical surfaces...... 18

Figure 2.14: Shape analysis flowchart...... 20

Figure 2.15: Processing flowchart showing the mandatory steps that must be performed for sulcal morphometry...... 22

xi List of figures

Chapter 3:

Figure 3.1: A) Left and right total thalamus volume of the 46 adolescent males with early- onset psychosis and the 34 adolescent male healthy controls. B) Left and right total thalamus volume of 13 adolescent males with schizophrenia and 17 adolescent males with bipolar I disorder. Note: Values are in cm3, the error bars represent 1 standard deviation. Thalamus volumes were obtained using FreeSurfer software...... 40

Figure 3.2: A) Left and right thalamus regional volumetric differences before correction for multiple comparisons between 34 adolescent male controls and 46 adolescent males with early-onset psychosis. B) and C) Right-sided thalamus regional volumetric differences after correction for multiple comparisons between 34 adolescent male controls and 46 adolescent males with early-onset psychosis. Note: The color bars show the magnitude and direction of the differences (distance in mm) between the average surfaces of each group obtained by SPHARM-PDM. A positive distance means that the average surface of the individuals in the early-onset psychosis group represents contraction with respect to the average surface of the controls (i.e., surface deflation of those in psychosis group) and vice versa...... 41

Figure 3.A.1: Left and right total thalamus volume of the 34 adolescent male healthy controls and 49 adolescent males with early-onset psychosis. Note: The error bars represent 1 standard deviation. Thalamus volumes were obtained using FreeSurfer (FS) and FSL-FIRST (FSL) software...... 47

Figure 3.A.2: A regression plot showing total thalamus volume from FreeSurfer vs. FSL- FIRST (FSL) data-sets, along with the correlation coefficient and regression equation relating the two sets of volumes...... 47

Figure 3.B.1: Box plots showing that increasing the SPHARM degree from 7 to 12 demonstrably shortens the mean distance, suggesting that increasing the number of SPHARM basis functions will result in a better approximation of the initial (not SPHARM) surface...... 49

Chapter 4:

Figure 4.1: Schematic representation of a gyrus and sulcus representing the different cortical morphometrics used in the current study...... 61

Figure 4.2: Parcellated hemispheric cortical surface overlaid with parcellated wire-frame representation of the cortical surface hull...... 62

Figure 4.3: Schematic representation of the image processing. FreeSurfer (v5.1) and BrainVISA (v4.2.1) software were combined...... 63

Figure 4.4: A) Pearson partial correlations between percentages change over time relative to baseline of the cortical morphological measures. CT, cortical thickness, SA, pial surface area, HS, hull surface area, WT, gyral WM thickness, SD, sulcal depth, SL, sulcal length, SW, sulcal width. Only partial correlations that were significant (p<0.05, two-tailed) after

xii List of figures

FDR correction (q = 0.05) are displayed. B, C, D, E and F) Scatter plots showing the relationship between different pairs of measures...... 67

Figure 4.B.1: Left column: Plots of longitudinal change in 52 healthy adolescents for lobar cortical thickness, pial surface area, gyrification index, hull surface area, gyral WM thickness, sulcal depth, length, and width. Values are raw values, averaged or summed (pial surface area, hull surface area) over lobe. Right column: Change over time relative to baseline for the same morphometric measures. Different colors indicate different sites and the p-value represents the effect of site on change in the morphometrical variable...... 75

Chapter 5:

Figure 5.1: The effect of site on the measurements. For each measure the unstandardized residuals (age, sex, site, ICV and handedness regressed out) were plotted by lobe. Subjects were color coded for site. The conglomeration of colors for all measures, i.e., the lack of a clear grouping of colors, suggests that when comparing the residuals no effect of site was present...... 88

Figure 5.2: Scatterplots showing significant differences after correction for multiple comparisons in cortical morphological measurements between healthy controls (n=52), early-onset psychosis (EOP) patients with a two-year follow-up diagnosis of schizophrenia (n=20) and EOP patients with a 2-year follow-up diagnosis of bipolar disorder with psychotic symptoms (n=20). Values are adjusted for age, sex, site, handedness and ICV, solid bar represents the median. C=controls; SCZ=schizophrenia; BD=bipolar disorder.92

Figure 5.3: Scatter plot of the relationship between sulcal width and the GI of the frontal cortex in the combined group of patients (r=-0.58, p<0.001) showing that an increased sulcal width was associated with a decreased GI. Solid line represents linear regression line...... 94

Figure 5.A.1: Quality control process to close the final study sample (green rectangle). Clinical or technical exclusion criterions as well as the number of excluded subjects are also displayed...... 98

Chapter 6:

Figure 6.A.1: Schematic representation of hardware organization...... 108

Figure 6.A.2: Schematic representation of CIBERSAM image processing platform...... 109

xiii

Motivation and objectives

Chapter 1 Motivation and objectives

1.1. Motivation

Due to its high structural and functional complexity, the human brain is the less known organ of the human body. Brain morphology is in continuous change across life span and its biological and functional implications remain unclear (Chi et al. 1977; Raznahan et al. 2011). Physicians and scientists, over many years, have traditionally assessed brain changes from a macroscopic point of view. Besides functional and cognitive information, structural neuroimaging has become a key technique for the evaluation and quantification, in vivo, of brain maturation changes related to neurodevelopment, aging, learning or disease (Rakic et al. 1994; Zilles et al. 1997; Good et al. 2001; Brown et al. 2012; van Soelen et al. 2012). Within the domain of medical imaging, magnetic resonance imaging (MRI) is a prominent technique that enables investigating complex spatial and functional relationships between different brain areas. The evolution of MRI sequences (Le Bihan 1995), linked to improvements in spatial resolution and image quality, has increased the knowledge about human brain neurodevelopment. Alongside medical image techniques, methods that employ magnetic resonance imaging information to convert the images into quantitative biomarkers such as size or shape have also been developed. These groups of methods and algorithms are known as morphometric techniques. Features such as volume, cortical thickness, surface area or cortical gyrification can be mapped within brain volume or onto brain surface, providing an ideal framework to assess their pattern and extent over time, across individuals or groups (Sowell et al. 2004; Im et al. 2008; Schaer et al. 2008; Shaw et al. 2008; Kochunov et al. 2010; Raznahan et al. 2011; Germanaud et al. 2014). Morphometric methods can be used for localizing significant structural differences between populations, or for showing that overall brain structure may be related to some effect of interest (Ashburner and Friston 2000). Initially, cross- sectional morphometric studies established relationships between anatomical biomarkers and demographic or clinical variables such as age, sex, cognitive ability,

1 Chapter 1 medication or disease. The benefit of a cross-sectional study design is that it allows researchers to compare many different variables at the same time. However, cross- sectional studies may not provide definite information about cause-and-effect relationships or change over time of desired biomarkers. Thus, longitudinal studies have been proposed to assess, at both group and individual level, the change over time of morphometric features (Raznahan et al. 2011) . According to the existence or not of a priori hypotheses about the spatial localization of the results, both cross-sectional and longitudinal morphometric studies can be categorized as: 1) region of interest (ROI) or 2) whole brain analyses. Region of interest analysis was the first strategy applied to the study and quantification of brain anatomy using MRI (Gur et al. 1998; Cusack 2005; Chang et al. 2005). In this type of analysis, one or more biomarkers are computed within a clearly defined region, which has been previously chosen according to a priori hypotheses or previous results. Its main drawback is that the analysis is fully conditioned by the initial hypothesis, making this methodology blind to results and inferences in discarded regions, which could be also relevant. On the other hand, whole brain morphometry methodology is a hypothesis-free strategy which gives an even-handed and comprehensive assessment of anatomical differences throughout the whole brain anatomy (Fischl et al. 1999; Ashburner and Friston 2000; Ashburner and Friston 2001; Bookstein 2001; Kim et al. 2005; Smith et al. 2006; Fischl 2012). This approach requires a previous spatial alignment between subject‟s images or surfaces by some form of spatial normalization. Thus, its main drawbacks are its high dependency on the normalization step and the statistical problem of dealing with multiple comparisons. Previous studies have shown that the various cortical morphometric features are not independent (Pomarol-Clotet et al. 2010; Uludag and Roebroeck 2014). The interplay between morphological features probably reflects the dynamic nature of the living brain, changes in one morphological feature due to e.g. development or disease indirectly affect other morphological features as well. By acknowledging this interdependence of characteristics a more complete assessment of brain morphology emerges. However, the standard approach employed in brain morphometric studies consists of assessing a single or minimal number of morphometric features, thus disregarding the importance of combining different structural features, multimodal imaging and a varied number of processing techniques to create an accurate overview

2 Motivation and objectives about human brain anatomy. In these cases, processing strategies, and therefore the hypotheses, are technically limited by the usage of an available processing package applied over images with the same MRI modality, thus leading to an incomplete results interpretation or in the worst cases to miss-interpretations. This document constitutes a thesis in the field of computer science with specific application to computational neuroscience and brain morphometry. Biomarkers, techniques and processing strategies are combined throughout this dissertation to provide morphometric tools for studying subcortical and cortical changes during neurodevelopment. Specifically, it addresses the design, development and implementation of new approaches as well as the application of pre-existing morphometric methods for the study of thalamus and brain cortex morphology in patients with psychosis and healthy subjects.

1.2. Hypothesis and objectives

The main hypothesis of this thesis is that a comprehensive characterization of human brain anatomy, in both healthy subjects and patients, can be only achieved if advanced biomarkers, computed through different image processing techniques, are combined under multimodal approaches. Under this idea, the main goal of this thesis is to develop processing strategies to combine different advanced morphometric techniques in order to study and characterize human brain anatomy. This global aim can be divided into three general sub-objectives: 1. To apply shape analysis techniques for the quantification of whole and regional thalamic volume, and to test them in real clinical scenarios. 2. To combine different cortical measurements in order to obtain a complete characterization of normal brain cortex development during adolescence. 3. To employ, simultaneously, different lobar biomarkers in order to investigate the mechanisms underlying the decrease in cortical thickness and gyrification observed in patients with a disease onset during adolescence.

1.3. Outline

This thesis is divided in six chapters. The first chapter (Chapter 1) provides a brief summary of the work, the motivation and general hypothesis as well as the

3 Chapter 1 objectives of the work. Chapter 2 presents a literature review about the most employed algorithms and methodologies for human brain morphometry using T1- weighted magnetic resonance imaging. Chapters 3, 4 and 5, contain the bulk of the work developed for the accomplishment of this thesis. Chapter 3 deals with shape analysis techniques to investigate global and local thalamic volume deficits in patients with psychosis. Chapter 4 presents a novel morphometric analysis procedure applied to a longitudinal study in which different cortical biomarkers such as cortical thickness, surface area, sulcal depth or sulcal width are combined to quantify the cortical surface flattening process that takes place during adolescence. In Chapter 5, the strategy developed in the previous chapter is employed to compare different cortical biomarkers between healthy control and early onset of psychosis (EOP)- schizophrenia and EOP-bipolar subjects. Finally, Chapter 6 closes this dissertation presenting the main conclusions achieved from this work. A general discussion and recommendation for further works are also presented. Some appendices contain additional material that provides extra technical details on particular issues.

4 Motivation and objectives

Related references

Ashburner, J. and K. J. Friston (2000). "Voxel-based morphometry--the methods." Neuroimage 11(6 Pt 1): 805-821. Ashburner, J. and K. J. Friston (2001). "Why voxel-based morphometry should be used." Neuroimage 14(6): 1238-1243. Bookstein, F. L. (2001). "Voxel-based morphometry" should not be used with imperfectly registered images." Neuroimage 14(6): 1454-1462. Brouwer, R. M., R. C. Mandl, et al. (2012). "White matter development in early puberty: a longitudinal volumetric and diffusion tensor imaging twin study." PLoS One 7(4): e32316. Cusack, R. (2005). "The intraparietal sulcus and perceptual organization." Journal of Cognitive Neuroscience 17(4): 641-651. Chang, K., A. Karchemskiy, et al. (2005). "Reduced amygdalar gray matter volume in familial pediatric bipolar disorder." Journal of the American Academy of Child & Adolescent Psychiatry 44(6): 565-573. Chi, J. G., E. C. Dooling, et al. (1977). "Gyral development of the human brain." Annals of neurology 1(1): 86-93. Fischl, B. (2012). "FreeSurfer." Neuroimage. 62(2): 774-781. Fischl, B., M. I. Sereno, et al. (1999). "Cortical surface-based analysis. II: Inflation, flattening, and a surface-based coordinate system." Neuroimage. 9(2): 195-207. Germanaud, D., J. Lefevre, et al. (2014). "Simplified gyral pattern in severe developmental microcephalies? New insights from allometric modeling for spatial and spectral analysis of gyrification." Neuroimage 102 Pt 2: 317-331. Good, C. D., I. S. Johnsrude, et al. (2001). "A voxel-based morphometric study of ageing in 465 normal adult human brains." Neuroimage 14(1 Pt 1): 21-36. Gur, R. E., V. Maany, et al. (1998). "Subcortical MRI volumes in neuroleptic-naive and treated patients with schizophrenia." American Journal of Psychiatry 155(12): 1711-1717. Im, K., J. M. Lee, et al. (2008). "Sulcal morphology changes and their relationship with cortical thickness and gyral white matter volume in mild cognitive impairment and Alzheimer's disease." Neuroimage 43(1): 103-113. Kim, J. S., V. Singh, et al. (2005). "Automated 3-D extraction and evaluation of the inner and outer cortical surfaces using a Laplacian map and partial volume effect classification." Neuroimage 27(1): 210-221. Kochunov, P., D. C. Glahn, et al. (2010). "Genetics of primary cerebral gyrification: Heritability of length, depth and area of primary sulci in an extended pedigree of Papio baboons." Neuroimage 53(3): 1126-1134. Le Bihan, D. (1995). "Diffusion, perfusion and functional magnetic resonance imaging." Journal des Maladies Vasculaires 20(3): 203-214. Pomarol-Clotet, E., E. J. Canales-Rodriguez, et al. (2010). "Medial prefrontal cortex pathology in schizophrenia as revealed by convergent findings from multimodal imaging." Molecular Psychiatry 15(8): 823-830. Rakic, P., J. P. Bourgeois, et al. (1994). "Synaptic development of the cerebral cortex: implications for learning, memory, and mental illness." Progress in Brain Research 102: 227-243.

5 Chapter 1

Raznahan, A., P. Shaw, et al. (2011). "How does your cortex grow?" Journal of Neuroscience 31(19): 7174-7177. Schaer, M., M. B. Cuadra, et al. (2008). "A surface-based approach to quantify local cortical gyrification." IEEE Transactions on Medical Imaging 27(2): 161-170. Shaw, P., N. J. Kabani, et al. (2008). "Neurodevelopmental trajectories of the human cerebral cortex." Journal of Neuroscience 28(14): 3586-3594. Smith, S. M., M. Jenkinson, et al. (2006). "Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data." Neuroimage 31(4): 1487-1505. Sowell, E. R., P. M. Thompson, et al. (2004). "Longitudinal mapping of cortical thickness and brain growth in normal children." Journal of Neuroscience 24(38): 8223-8231. Uludag, K. and A. Roebroeck (2014). "General overview on the merits of multimodal neuroimaging data fusion." Neuroimage 102 Pt 1: 3-10. van Soelen, I. L., R. M. Brouwer, et al. (2012). "Genetic influences on thinning of the cerebral cortex during development." Neuroimage 59(4): 3871-3880. Zilles, K., A. Schleicher, et al. (1997). "Quantitative analysis of sulci in the human cerebral cortex: development, regional heterogeneity, gender difference, asymmetry, intersubject variability and cortical architecture." Human Brain Mapping 5(4): 218-221.

6 State of the art in brain morphometry

Chapter 2 State of the art in brain morphometry

Introduction

Brain morphometry is a research area where two major actual research lines converge: image processing and neuroscience. Its main goal is to obtain quantitative biomarkers that may enable an accurate characterization of brain anatomy. The evolution of neuroimaging modalities, especially magnetic resonance imaging (MRI), has allowed researchers to quantify, in vivo, changes in brain morphology related to neurodevelopment, aging, learning or disease. The most frequently employed MRI modalities for brain morphometry are conventional structural MRI (T1 and T2- weighted) MRI and diffusion weighted imaging (DWI). This chapter presents a brief review focusing on the most popular and commonly used methods for brain T1-weighted MRI morphometry, as well as some important image preprocessing concepts used to obtain an accurate structural characterization of human brain.

2.1. Preprocessing steps in brain morphometry

Morphometric techniques need a number of initial standard preprocessing steps such as image enhancement, tissue segmentation or surface extraction before subsequent processing steps can be applied. This section contains a brief introduction to some of these initial preprocessing steps.

2.1.1. Inhomogeneity correction

One of the greatest problems faced by MRI processing techniques is nonuniformity of the images intensity (see Figure 2.1). Inhomogeneity artifacts introduce smooth variations in the intensity of the images, causing that regions belonging to the same tissue type present different intensity levels. Many research groups have designed methods for image enhancement by estimating a smooth function that modulates the image intensities (Sled et al. 1998; Hou et al. 2006; Hui et al. 2010).

7 Chapter 2

Figure 2.1: Inhomogeneity artifact in a T1-weighted MRI.

Correction of the non-uniformity artifact implies that different tissues retain their intrinsic intensity pattern throughout the image, thus reducing possible errors in future pre-processing steps such as image registration or tissue segmentation.

2.1.2. Image registration and spatial normalization

Image registration is defined as the process of applying geometric transformations to the space wherein a volumetric image is defined such that it is perfectly aligned with another volume, following some similarity criteria. There are several criteria for classifying registration algorithms but the most extended one is based on the geometrical deformation applied during the registration process. Two basic types of deformations can be applied during the registration process: linear and nonlinear transformations (Figure 2.2). Linear transformations, also known as affine transformations, act globally over the images and include rigid body transformations (rotation and translations) as well as scaling, shearing and perspective deformations (Ashburner et al. 1997). The main advantage of linear registration methods is their low computation time due to the fact that only fifteen parameters need to be estimated.

8 State of the art in brain morphometry

Figure 2.2: Linear (affine) transformations. A) Translation. B) Rotation. C) Scaling. D) Shearing. E) Perspective deformation.

The main drawback of linear registration methods is that they are not capable of obtaining local deformations leading to an inexact anatomic correspondence between images from different subjects due to the high local variability of human brain morphology. Nonlinear registration methods are intended to reduce such local differences (Figure 2.3). Nonlinear alignment obtains a deformation field which can be considered as a continuous 3D vector field. Each transformation vector represents the displacement to be applied to an image point to reach its homologous point on a target image. For this reason, these methods present a higher number of degrees of freedom (number of parameters needed to describe the registration transformation) than affine registration methods, making the estimation more complex and unstable.

Commonly, the application of constraints over the deformation field makes the nonlinear registration computationally tractable and physical plausible. In studies of soft tissue deformation, understanding the physics of the deformation process might reduce

9 Chapter 2 the number of degrees of freedom of the transformation. Models based on the laws of physics have been used to reduce the degrees of freedom of the matching process in inter-subject registration (Toga 1999). Some widespread used registration algorithms are: HAMMER (Hierarchical Attribute Matching Mechanism for Elastic Registration) (Shen and Davatzikos 2002), DARTEL (Diffeomorphic Anatomical Registration using Exponentiated Lie algebra) (Ashburner 2007) and ANTs (Advance Normalization Tools) (Avants et al. 2006; Avants et al. 2008). A large evaluation of some of these nonlinear registration algorithms was presented by Klein and colleagues in (Klein et al. 2009).

Spatial normalization Spatial normalization is defined as the registration of one or more images to a common, also named stereotactic, space. This common geometrical space can be defined by a single image or by the average of a previously registered group of images. Spatial normalization is employed to superimpose common regions, statistically analyze the signal in them and make inferences about brain structure and function in a specific framework.

2.1.3. Tissue classification

Human brain comprises different tissues; the principal ones are white matter (WM), grey matter (GM) and cerebrospinal fluid (CSF) (see Figure 2.4). The white matter of the brain is made up primarily of axon tracts, the long, spindly appendages of neurons. These tracts transmit the electrical signals that the neurons, comprising the grey matter tissue, use to communicate with each other. They are wrapped (to a different degree depending on the particular tract) in a fatty layer called myelin, which insulates the axons and allows for rapid signal conduction.

Figure 2.4: Brain tissues segmentation from T1-weighted images. Red: Cerebrospinal fluid. Magenta: White matter. Green: Grey matter.

10 State of the art in brain morphometry

Depending on the imaging parameters of the MR acquisition, different tissues yield varying signal responses, which are used to discern tissue type distribution in the image. Tissue segmentation algorithms employ a combination of image intensity and spatial localization patterns in order to extract and classify brain tissues from MRIs. Some of the most employed brain segmentation algorithms were proposed by (Zhang et al. 2001; Ashburner and Friston 2005; Rivest-Henault and Cheriet 2011; Foruzan et al. 2013; Somasundaram and Kalavathi 2013).

2.1.4. Structure parcellation

Brain structure parcellation is a particular case of tissue segmentation. Once the tissues are correctly segmented, the next step is to determine which regions, inside the same tissue, belong to the same neuroanatomical region. Structure parcellation can be performed by manual, automatic or semi-automatic methods (Collins and Evans 1999; Fischl et al. 2002; Alemán-Gómez et al. 2006) (see Figure 2.5). Structure parcellation is particularly important in applications such as: extraction of anatomical regions of interest, volumetric measurements, 3D visualization, planning of radiotherapy or surgery planning.

Figure 2.5: Flowchart showing the steps followed by registration based algorithm for the brain structures parcellation.

11 Chapter 2

2.1.5. Cortical surface extraction

Once white matter, grey matter and cerebrospinal fluid volumes have been generated during a tissue segmentation step, surface tessellations of grey/white (white surface) and grey/CSF (pial surface) boundaries can be calculated (see Figure 2.6). Each surface is modeled as a mesh comprised by vertices (X, Y and Z coordinates obtained during the surface extraction process from the MRI) and triangles known as surface faces (Davatzikos and Bryan 1996; Dale et al. 1999; Kim et al. 2005).

Figure 2.6: Pial and white surfaces extracted from T1-weighted MRI for both hemispheres. Note: Pial surface is the GM/CSF boundary and white surface is the WM/GM boundary.

Another important surface that can be reconstructed is the hull surface. The hull surface is a smooth envelope wrapped around the pial surface that does not encroach into the sulci (Figure 2.7). It is important to note that the hull surface is different from the inflated surface. The inflated surface is used for improving visualization; e.g. displaying cortical areas that are buried inside sulci (see Figure 2.8, second row). Along the cortical surfaces, both global and local cortical morphometric measures can be computed. Some of the most important global indices are the surface area (SA) and the gyrification index (GI). The surface area is obtained by summing up the areas of the pial tessellation triangles and the GI is defined by the ratio between the area of the pial and hull surfaces (Figure 2.7) (Van Essen et al. 2001). GI quantifies the amount of cortex buried within the sulcal folds as compared with the amount of outer visible cortex. It was initially calculated in 2D as the ratio of the perimeter of the total cortical outline to that of the exposed brain surface, traced on 2D histological sections (Zilles et al. 1988; Zilles et al. 1989).

12 State of the art in brain morphometry

Figure 2.7: Hemispheric pial and hull surfaces. The regional gyrification index is the quotient between the area of the pial and hull surfaces.

Local indices can be also obtained for each surface vertex; the cortical thickness (CT) (Figure 2.8, column B), the curvature (CU) (Figure 2.8, column C) and the local gyrification index (lGI) (Figure 2.8, column D) being the most important vertex-wise features.

13 Chapter 2

The cortical thickness, for a certain point, is the distance between the white surface and its corresponding point on the pial surface. Cortical thickness is a direct measure of the amount of grey matter contained in cortical layers along the perpendicular direction to each point on the surface. The curvature map (Davatzikos et al. 2001) stores spatial information about gyral (convex) and sulcal (concave) regions over the surface. On the other hand, similar to the global gyrification index, the local gyrification index (LGI) (Schaer et al. 2008) brings information about cortical folding in a given neighborhood around each surface vertex.

2.1.6. Volumetric and geodesic smoothing

Another important step in image preprocessing is smoothing. The main purpose of smoothing is to increase the signal-to-noise ratio by reducing the high-frequency random noise. During the smoothing process, the value assigned to each voxel (Figure 2.9A) or vertex (Figure 2.9B) is replaced by a weighted average of the intensity of its surrounding neighbors. The spatial extension of the neighborhood is determined by the size of the smoothing kernel, which can vary across studies. Smoothing makes the data conform more closely to the Gaussian field model, increasing the validity of parametric tests (Ashburner and Friston 2000). Inaccurate registration or human neuroanatomical variability leads to increased variance which may reduce the power to detect the signal of interest. Smoothing also serves to reduce this variance. Although, excessive smoothing will decrease the capability for accurately localizing changes in brain morphology (Ashburner and Friston 2000).

Figure 2.9: Smoothing. A) Volumetric smoothing. B) Geodesic smoothing.

14 State of the art in brain morphometry

2.2. Region of interest analysis

The previously described preprocessing steps are necessary for quantification of brain morphology. The quantification of brain anatomy using MRI used to be done with region of interest (ROI) analyses. In ROI analyses a number of steps have to be taken. The voxels or vertices that belong to a specific ROI have to be selected, from volumetric image or surface. The choice of the ROI is usually done based on an a priori hypothesis. Once the ROI is selected, different morphological features, such as volume, surface or shape descriptors, can be obtained and employed in subjects or group comparisons. The main advantage of a ROI analysis is that the analysis can be performed in the individual subject space, avoiding issues with interpolation and miss-registration that can occur during the normalization process. Its main drawback is that the analysis is fully conditioned by the initial hypothesis. Thus, this methodology precludes results and inferences about discarded regions that may be of importance. To avoid the conditioning of results by ROI selection whole brain morphometric methods have been developed. The main advantage of whole brain morphometric methods is that they are not focused on a priori selected regions and give an even-handed and comprehensive assessment of anatomical differences throughout the brain (Ashburner and Friston 2000). The most important whole brain morphometric methods are deformation based morphometry (DBM) (Ashburner et al. 1998), tensor based morphometry (TBM) (Hua et al. 2013) and the most known and employed: voxel based morphometry (VBM) (Ashburner and Friston 2000) and surface based morphometry (SBM).

2.3. Voxel based morphometry

Voxel based morphometry (VBM) is a neuroimaging analysis technique that performs statistical test across all voxels in the image to identify volume differences between groups. For VBM, different preprocessing steps need to be performed before the statistical analysis (see Figure 2.10). Firstly, images are segmented into different tissue compartments (grey matter, white matter, and cerebrospinal fluid). VBM analysis is performed separately on either grey or white matter, dependent on the question being asked.

15 Chapter 2

Figure 2.10: Voxel based morphometry flowchart.

Secondly, tissues segmentations are normalized to a standard space to remove the inter-subject variability. Different algorithms can be used for performing the image normalization step (Ashburner and Friston 2000; Ashburner and Friston 2001) and a specific modality template or a customized sample template can be created to increase the registration accuracy (Good et al. 2001; Senjem et al. 2005). VBM relies on the assumption that the normalization step guarantees a spatial correspondence between anatomical regions. This assumption does not hold in regions with high anatomical variability between subjects such as prefrontal gyral or sulcal regions. Thirdly, a step called modulation may be applied to correct for tissue volume changes during the spatial normalization step (Good et al. 2001). Voxel intensities are scaled by the amount of expansion or contraction that has occurred during spatial normalization, so that the total amount of “tissue” remains the same as in the original image. Recent studies assessing the effects of modulation on results from VBM have questioned the modulation step, claiming that this step could decrease the sensitivity of the VBM method (Radua et al. 2014). Finally, tissue segmentation images are smoothed and fed into the statistical analysis. General linear model (parametric statistics) or nonparametric testing (Nichols and Holmes 2002; Rorden et al. 2007) can be used to analyze the data. VBM generates maps displaying the voxels where a statistically significant effect is found. Although both grey and white matter volumes can be assessed using VBM, the majority of VBM studies focus on grey matter. Changes in white matter are more

16 State of the art in brain morphometry accurately assessed using parametric images obtained from diffusion weighted imaging such as fractional anisotropy or mean diffusivity maps (Smith et al. 2006; Melonakos et al. 2011). The main advantage of VBM is that it can perform the analysis over the entire brain volume without the requirement of establishing a prior spatial hypothesis. Its main drawback is its high dependency on the preprocessing steps (degree of smoothing, tissue segmentation method (Figure 2.11B) and especially the registration step (Figure 2.11C)). Another major limitation is the problem with interpreting VBM results as this method is not capable of distinguishing whether the observed volume change is related to a change in cortical thickness (Figure 2.11D) or due to a change in the cortical surface area (Figure 2.11E).

Figure 2.11: Possible causes of the differences obtained during a VBM analysis.

2.4. Surface based morphometry or vertex-wise analysis

Surface based morphometry (SBM) methods solve the existing uncertainty of VBM result interpretation. SBM decomposes cortical volume into its two mathematically independent and physically defined constituents: surface area and cortical thickness (Figure 2.12).

Figure 2.12: Volume as a composite measure of surface area and cortical thickness.

Similar to the VBM stream, the input for SBM is a high-resolution T1-weighted MRI but in this case the statistical analysis is performed by using morphometric features obtained from geometric models of the cortical surface.

17 Chapter 2

There are several implementations of SBM (Fischl et al. 1999; Kim et al. 2005) but most of them contain similar processing steps. The first steps of SBM include cortical surface extraction and calculation of vertex-wise morphological features such as cortical thickness, surface area, curvature and local gyrification index. Thereafter, the white surface is mapped to a unit sphere (Fischl et al. 1999) and the curvature map is used for surface-based spatial normalization (Fischl et al. 1999). During this process, the 2D spherical coordinate of a vertex is adjusted so as to best match the curvature across subjects. Aligning the curvature aligns the folding patterns (i.e., the gyri and sulci). The main argument for this is that cortical folding patterns are robust neuroanatomical landmarks for registration (Fischl et al. 1999) (Figure 2.13). Spatial normalization in SBM methods causes that parts of the surface may be compressed or stretched in order to obtain the best match between the curvature maps. The obtained spatial transformation is applied to each subject‟s cortical thickness map to re-map it into a common surface group space. This step allows for comparing surface maps across subjects at homologous points along the cortex.

Figure 2.13: Surface alignment and normalization strategy for cortical surfaces.

18 State of the art in brain morphometry

Finally, normalized surface maps are geodesically smoothed and fed into a group analysis. Similar to VBM, SBM generates statistical maps showing vertices where a statistically significant effect for cortical thickness, local gyrification index, etc. between groups was found. The main advantages of SBM methods are: 1. They can perform the analysis over the entire cortical surface simultaneously. 2. The curvature-based registration step increases the vertex correspondence accuracy in gyral and sulcal regions. The accurate correspondence between homologous anatomical regions increases the sensibility of the method. 3. SBM results can distinguish whether volume findings are due to differences in cortical thickness or surface area. 4. SBM results interpretation is easier than VBM because the analysis involves parameters in units of measurement (cortical thickness, cortical surface area, curvature, etc). Its main drawback is the high dependency on the surface extraction step. Topological errors in white and pial surfaces can lead to estimation errors in surface maps and can also affect the spherical mapping. Although the results interpretation is more straightforward than VBM, the biological interpretation of the SBM results remains complicated due to the macroscopic nature of the measurements. For example, cortical thinning during adolescence could reflect synaptic pruning or increased myelination (Paus et al. 2008), whereas it is associated with neuronal loss in neurodegenerative disease (Dickerson et al. 2009).

2.5. Structural shape analysis

Despite being enclosed within SBM, shape analysis technique is used to assess surface shape changes of less spatially complex regions such as subcortical brain structures: the thalamus, amygdala, and hippocampus. Quantitative morphologic characterization of individual brain structures is often based on volumetric measurements. Volume changes are larger scale intuitive features as they might explain atrophy or dilation due to illness but fine-grained structural changes are not sufficiently reflected in volume measurements.

19 Chapter 2

Shape analysis techniques are becoming increasingly popular to neuroscientists due to their capacity for precisely locating subcortical morphological changes between healthy and pathological structures (Davatzikos et al. 1996; Joshi et al. 1997; Csernansky et al. 2002; Styner et al. 2006). The first step for shape analysis is a parcellation that accurately delineates the brain structure of interest. Secondly, the volumetric structure segmentation is converted to a 3D mask boundary surface mesh by using a tessellation algorithm. Thirdly, the surface mesh is mapped to a spherical coordinate system through a SPHARM-based (Styner et al. 2006) or Wavelets-based (Yu et al. 2007) shape parameterization. The surface is characterized in the spherical coordinate system by sets of coefficients weighting the SPHARM/Wavelets basis functions. Fourthly, the SPHARM/Wavelets-based surface is aligned and subdivided into an equal number of surface points via an icosahedron subdivision level to establish spatial correspondence across them. Although each surface point has a one-to-one correspondence across subjects, they remain in the native image space. Therefore, the surfaces of all study participants must be aligned to a common space prior to any statistical analysis. Alignment is achieved by computing a mean template of the structure‟s surface over all participants. Finally, the point-wise distance map between each individual‟s surface and the reference surface is used for the statistical analysis (Figure 2.14).

Figure 2.14: Shape analysis flowchart.

20 State of the art in brain morphometry

Given a smooth and fine-scale shape representation, the shape analysis method enables an accurate subregional volume assessment but two main limitations need to be taken into account: The structure boundary strongly depends on the surface extraction algorithm. Different surface extraction algorithms and surface subdivision levels lead to different surface boundaries. The spherical harmonic or wavelets series truncation at different degrees also lead to different representations at different levels of detail of the surface.

2.6. Sulcal morphometry

The last strategy that will be described in this brief introduction to human brain morphometry methodologies is sulcal morphometry. Cortical sulci are structures of the brain which show strong morphological change due to brain development or pathology (Zilles et al. 1997; Rametti et al. 2010; Malikovic et al. 2012). Cortical surfaces are employed for reconstructing sulci structures in order to obtain sulci related biomarkers such as its depth, width or length. Sulci morphometry algorithms employs a crevasse detector to reconstruct sulcal surface as the medial surfaces from the two opposing gyral banks that span from the most internal point of sulcal fold to the hull of the cortex (Figure 2.15) (Mangin et al. 2004). Thereafter, neural network-based pattern classifiers, atlas based or manual labeling could be employed for automatic identification of the cortical sulci. Finally, for each sulcus, its depth, length and width as biomarkers can be computed. This processing pipeline is implemented in BrainVISA (www.brainvisa.info), which is the most used image processing package for sulci morphometry. Sulci morphometry contains a large number of processing steps (Figure 2.15), all of them with their own drawbacks. Because of substantial variability of the sulci in shape and size, labeling is the step with a greater impact on the quality of the results. Wrong labeling leads to wrong values for sulci depth, length and width measurements.

21 Chapter 2

Figure 2.15: Processing flowchart showing the mandatory steps that must be performed for sulcal morphometry.

On the other hand, sulci morphometry allows for observing and quantifying the influence of the cerebral sulcal anatomy over cortical shape, providing the neuroscientists with the tools to create a more complete picture of cortical morphology. Sulcal morphometry and previously presented processing strategies are combined in the next chapters to obtain a comprehensive and accurate characterization of human brain anatomy in both healthy subjects and patients.

22 State of the art in brain morphometry

Related references

Alemán-Gómez, Y., L. Melie-García, et al. (2006). "IBASPM: Toolbox for automatic parcellation of brain structures." Organization for Human Brain Mapping, Florence, Italy, NeuroImage. Ashburner, J. (2007). "A fast diffeomorphic image registration algorithm." Neuroimage 38(1): 95-113. Ashburner, J. and K. J. Friston (2000). "Voxel-based morphometry--the methods." Neuroimage 11(6 Pt 1): 805-821. Ashburner, J. and K. J. Friston (2001). "Why voxel-based morphometry should be used." Neuroimage 14(6): 1238-1243. Ashburner, J. and K. J. Friston (2005). "Unified segmentation." Neuroimage 26(3): 839-851. Ashburner, J., C. Hutton, et al. (1998). "Identifying global anatomical differences: deformation-based morphometry." Human Brain Mapping 6(5-6): 348-357. Ashburner, J., P. Neelin, et al. (1997). "Incorporating prior knowledge into image registration." Neuroimage 6(4): 344-352. Avants, B., C. Epstein, et al. (2008). "Symmetric diffeomorphic image registration with cross- correlation: evaluating." Medical Image Analysis 12(1): 26-41. Avants, B., P. T. Schoenemann, et al. (2006). "Lagrangian frame diffeomorphic image registration: Morphometric comparison of human and chimpanzee cortex." Medical Image Analysis 10(3): 397-412. Collins, L. and A. C. Evans (1999). "ANIMAL: Automatic non linear image matching for automatic labelling." Brain Warping, Academic Press. Csernansky, J. G., L. Wang, et al. (2002). "Hippocampal deformities in schizophrenia characterized by high dimensional brain mapping." American Journal of Psychiatry 159(12): 2000-2006. Dale, A. M., B. Fischl, et al. (1999). "Cortical surface-based analysis. I. Segmentation and surface reconstruction." Neuroimage. 9(2): 179-194. Davatzikos, C. and N. Bryan (1996). "Using a deformable surface model to obtain a shape representation of the cortex." IEEE Transactions on Medical Imaging. 15(6): 785-795. Davatzikos, C., A. Genc, et al. (2001). "Voxel-based morphometry using the RAVENS maps: methods and validation using simulated longitudinal atrophy." Neuroimage 14(6): 1361-1369. Davatzikos, C., M. Vaillant, et al. (1996). "A computerized method for morphological analysis of the corpus callosum." Journal of Computed Assisted Tomography 20: 88-97. Dickerson, B. C., A. Bakkour, et al. (2009). "The cortical signature of Alzheimer's disease: regionally specific cortical thinning relates to symptom severity in very mild to mild AD dementia and is detectable in asymptomatic amyloid-positive individuals." Cerebral Cortex 19(3): 497-510. Fischl, B., D. H. Salat, et al. (2002). "Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain." Neuron 33(3): 341-355. Fischl, B., M. I. Sereno, et al. (1999). "Cortical surface-based analysis. II: Inflation, flattening, and a surface-based coordinate system." Neuroimage. 9(2): 195-207.

23 Chapter 2

Foruzan, A. H., I. Kalantari Khandani, et al. (2013). "Segmentation of brain tissues using a 3-D multi-layer hidden Markov model." Computers in Biology and Medicine. 43(2): 121- 130. Good, C. D., I. S. Johnsrude, et al. (2001). "A voxel-based morphometric study of ageing in 465 normal adult human brains." Neuroimage 14(1 Pt 1): 21-36. Hellier, P. and C. Barillot (2003). "Coupling dense and landmark-based approaches for nonrigid registration." IEEE Transactions on Medical Imaging 22(2): 217-227. Hou, Z., S. Huang, et al. (2006). "A fast and automatic method to correct intensity inhomogeneity in MR brain." Medical Image Computing and Computer Assisted Intervention 9(Pt 2): 324-331. Hua, X., D. P. Hibar, et al. (2013). "Unbiased tensor-based morphometry: improved robustness and sample size estimates for Alzheimer's disease clinical trials." Neuroimage 66: 648- 661. Hui, C., Y. X. Zhou, et al. (2010). "Fast algorithm for calculation of inhomogeneity gradient in magnetic resonance." Journal of Magnetic Resonance Imaging 32(5): 1197-1208 Joshi, S. C., M. I. Miller, et al. (1997). "On the Geometry and Shape of Brain Sub-Manifolds." International Journal of Pattern Recognition and Artificial Intelligence 11: 1317-1343. Kim, J. S., V. Singh, et al. (2005). "Automated 3-D extraction and evaluation of the inner and outer cortical surfaces using a Laplacian map and partial volume effect classification." Neuroimage. 27(1): 210-221. Klein, A., J. Andersson, et al. (2009). "Evaluation of 14 nonlinear deformation algorithms applied to human brain MRI." Neuroimage 46(3): 786-802 Le Goualher, G., E. Procyk, et al. (1999). "Automated extraction and variability analysis of sulcal neuroanatomy." IEEE Transactions on Medical Imaging 18(3): 206-217. Malikovic, A., B. Vucetic, et al. (2012). "Occipital sulci of the human brain: variability and morphometry." Anatomical Science International 87(2): 61-70. Mangin, J. F., D. Riviere, et al. (2004). "Object-based morphometry of the cerebral cortex." IEEE Transactions on Medical Imaging 23(8): 968-982. Melonakos, E. D., M. E. Shenton, et al. (2011). "Voxel-based morphometry (VBM) studies in schizophrenia-can white matter changes be reliably detected with VBM?" Psychiatry Research 193(2): 65-70. Nichols, T. E. and A. P. Holmes (2002). "Nonparametric permutation tests for functional neuroimaging: a primer with examples." Human Brain Mapping 15(1): 1-25. Paus, T., M. Keshavan, et al. (2008). "Why do many psychiatric disorders emerge during adolescence?" Nature Reviews Neuroscience. 9(12): 947-957. Perrot, M., D. Riviere, et al. (2011). "Cortical sulci recognition and spatial normalization." Medical Image Analysis 15(4): 529-550. Radua, J., E. J. Canales-Rodriguez, et al. (2014). "Validity of modulation and optimal settings for advanced voxel-based morphometry." Neuroimage 86: 81-90. Rametti, G., C. Junque, et al. (2010). "Anterior cingulate and paracingulate sulci morphology in patients with schizophrenia." Schizophrenia Research 121(1-3): 66-74. Rivest-Henault, D. and M. Cheriet (2011). "Unsupervised MRI segmentation of brain tissues using a local linear model and level set." Magn Reson Imaging 29(2): 243-259. Rorden, C., L. Bonilha, et al. (2007). "Rank-order versus mean based statistics for neuroimaging." Neuroimage 35(4): 1531-1537.

24 State of the art in brain morphometry

Rosen, H. J., M. L. Gorno-Tempini, et al. (2002). "Patterns of brain atrophy in frontotemporal dementia and semantic dementia." Neurology 58(2): 198-208. Schaer, M., M. B. Cuadra, et al. (2008). "A surface-based approach to quantify local cortical gyrification." IEEE Transactions on Medical Imaging 27(2): 161-170. Senjem, M. L., J. L. Gunter, et al. (2005). "Comparison of different methodological implementations of voxel-based morphometry in neurodegenerative disease." Neuroimage 26(2): 600-608. Shen, D. and C. Davatzikos (2002). "HAMMER: hierarchical attribute matching mechanism for elastic registration." IEEE Transactions on Medical Imaging 21(11): 1421-1439. Sled, J., A. Zijdenbos, et al. (1998). "A nonparametric method for automatic correction of intensity nonuniformity in MRI." IEEE Transactions on Medical Imaging 17(1): 87-97. Smith, S. M., M. Jenkinson, et al. (2006). "Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data." Neuroimage 31(4): 1487-1505. Somasundaram, K. and P. Kalavathi (2013). "Contour-based brain segmentation method for magnetic resonance imaging human head." J Comput Assist Tomogr 37(3): 353-368. Styner, M., I. Oguz, et al. (2006). "Framework for the Statistical Shape Analysis of Brain Structures using SPHARM-PDM." Insight Journal (1071): 242-250. Toga, A. (1999). "Brain Warping." Academic Press.

Van Essen, D. C., H. A. Drury, et al. (2001). "An integrated software suite for surface-based analyses of cerebral cortex." Journal of the American Medical Informatics Association 8(5): 443-459. Whitwell, J. L., C. R. Jack, Jr., et al. (2009). "Voxel-based morphometry patterns of atrophy in FTLD with mutations in MAPT or PGRN." Neurology 72(9): 813-820. Yu, P., P. E. Grant, et al. (2007). "Cortical surface shape analysis based on spherical wavelets." IEEE Transactions on Medical Imaging 26(4): 582-597. Zhang, Y., M. Brady, et al. (2001). "Segmentation of brain MR images through a hidden Markov random field model and the expectation-maximization algorithm." IEEE Transactions on Medical Imaging. 20(1): 45-57. Zilles, K., E. Armstrong, et al. (1989). "Gyrification in the cerebral cortex of primates." Brain, Behavior and Evolution 34(3): 143-150. Zilles, K., E. Armstrong, et al. (1988). "The human pattern of gyrification in the cerebral cortex." Anatomy and Embryology (Berl). 179(2): 173-179. Zilles, K., A. Schleicher, et al. (1997). "Quantitative analysis of sulci in the human cerebral cortex: development, regional heterogeneity, gender difference, asymmetry, intersubject variability and cortical architecture." Human Brain Mapping 5(4): 218-221.

Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis

Chapter 3 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis

Joost Janssen, Yasser Alemán-Gómez, Santiago Reig, Hugo G. Schnack, Mara Parellada, Montserrat Graell, Carmen Moreno, Dolores Moreno, J. M. Mateos-Pérez, J. M. Udias, Celso Arango and Manuel Desco

British Journal of Psychiatry. 2012 Jan;200(1):30-6. doi: 10.1192/bjp.bp.111.093732.

Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis

Abstract

Thalamic volume deficits are associated with psychosis but it is unclear whether the volume reduction is uniformly distributed or if it‟s more severe in particular thalamic regions. The main goal of this work is to quantify whole and regional thalamic volume in male early-onset psychosis (EOP) patients and male healthy comparison subjects. Brain scans were obtained for 83 adolescents (less than 18 years of age), 49 EOP patients (duration of positive symptoms less than 6 months) and 34 healthy controls. Total thalamic volumes were assessed using FreeSurfer and FSL-FIRST, regional thalamic volumes were examined with a surface-based approach. Total thalamic volume was smaller in EOP patients relative to controls. Regional thalamic volume reduction was most significant in the right anterior-mediodorsal area and pulvinar. In minimally treated male EOP patients, thalamic volume deficits may be most pronounced in the anterior-mediodorsal and posterior pulvinar regions, adding strength to findings from post mortem studies in adults with psychosis.

29 Chapter 3

3.1. Introduction

Chronic adult-onset schizophrenia is associated with smaller whole thalamic volume (Adriano et al. 2010) and post-mortem findings point to selective thalamic deficits in the anterior and mediodorsal nuclei and the pulvinar (Byne et al. 2008). The thalamus is a relay station that contains abundant efferent and afferent connections to the cortex. Therefore, structural alterations to the thalamus can cause a disruption in the thalamic–cortical–thalamic circuitry, leading to the typical psychosis- like cognitive and clinical symptoms (Jones 1997). A recent meta-analysis concluded that few studies investigated thalamus deficits in individuals with first-episode psychosis (Adriano et al. 2010). Given that antipsychotic medication may alter thalamic volume (Gur et al. 1998) it makes sense to measure thalamic volume in individuals with first-episode psychosis, particularly in adolescents with a short duration of illness and low doses of antipsychotic medication. Furthermore, studying these early-onset groups provides information about whether the thalamic abnormalities seen in adolescents are consistent with or different from the „adult‟ expression of these disorders (Frazier et al. 2008). Early-onset psychosis will convert into different diagnostic categories, such as adolescent-onset schizophrenia and early-onset bipolar I disorder, over time. These diagnoses may share some degree of underlying pathology as evidenced by common volume deficits in cortical structures such as the insular, anterior cingulate and orbitofrontal cortex (Janssen et al. 2008) and subcortical structures such as the hippocampus (Frazier et al. 2005). A smaller thalamic volume has been found in early- onset schizophrenia (Kumra et al. 2000) although the results for early-onset bipolar disorder are inconsistent (Dasari et al. 1999; Terry et al. 2009). In adolescents with psychosis it is unclear whether volumetric thalamic abnormalities are specific to particular thalamic nuclei as opposed to affecting the entire thalamus uniformly. In this study a total and regional thalamus volume between male adolescents (younger than 18 years of age) with early-onset first-episode psychosis (duration of positive symptoms less than 6 months) and healthy male controls was examined. This study sample eliminates the potentially confounding effects of gender.

30 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis

3.2. Materials and methods

3.2.1. Participants

The participants were 80 males: 46 with first-episode psychosis (psychosis group) and 34 healthy controls (control group). The psychosis group were recruited from the two child and adolescent psychiatry in-patient units in Madrid (Hospital General Universitario Gregorio Marañón and Hospital Universitario Infantil Niño Jesús). These two units serve a population of approximately five million people. All males consecutively seen at these facilities between April 2002 and November 2005 who fulfilled the inclusion criteria described below were invited to participate in the study. At baseline, 58 individuals were eligible for the study. Seven of these refused the magnetic resonance imaging (MRI) scan because of fear. Furthermore, two individuals were excluded because of insufficient image quality for neuroimaging analyses. At follow-up, three participants were excluded because they no longer fulfilled diagnostic inclusion criteria for psychosis (see Clinical assessment), thus leaving a sample of 46 individuals. The inclusion criteria for the psychosis group were: male, presence of positive psychotic symptoms within a DSM-IV (APA 1994) psychotic disorder (including schizophrenia, bipolar I disorder, schizoaffective disorder, schizophreniform disorder, delusional disorder, shared psychotic disorder, brief psychotic disorder, psychotic disorder not otherwise specified (NOS), major depression with psychotic features) before the age of 18 years and a maximum 6-month history of positive symptomatology at the time of enrolment. Exclusion criteria were: presence of other concomitant Axis I disorder at the time of evaluation that might account for the psychotic symptoms (such as autism-spectrum disorders, post-traumatic stress disorder or acute stress disorder), mental retardation per the DSM-IV criteria, including not only an IQ score (Wechsler 1974; Wechsler 1997) below 70 but also impaired functioning, neurological disorders, and a history of head trauma with loss of consciousness. Individuals that fulfilled criteria for presence of substance misuse or dependence at baseline were excluded. All potential participants had a careful assessment of drug-use history and drug urine analysis at baseline. If the assessment was positive individuals were included in the study only if positive symptoms persisted 14 days after a negative urine analysis. Six individuals had a history of cannabis use at baseline, but the distribution of their

31 Chapter 3 thalamus volumes within the psychosis group was even (not biased to smaller or larger values). Urine test analyses were repeated at various time points during a 2-year clinical follow-up. All participants in the psychosis group had a thorough medical examination as part of the standard clinical guidelines protocol. The control group consisted of 34 healthy males recruited from the same schools and residential areas as the psychosis group. The inclusion criteria for controls were similar age and coming from the same geographical areas as those in the psychosis group, no current and previous psychiatric disorder as measured by the Kiddie Schedule for Affective Disorders and Schizophrenia, Present and Lifetime version (K-SADS-PL) (Kaufman et al. 1997) and no neurological disorders, or head trauma or mental retardation (per the DSM-IV criteria). Further information regarding the inclusion and exclusion criteria can be found in Castro-Fornieles et al.2007 (Castro-Fornieles et al. 2007). The study was approved by the institutional review boards of both participating clinical centres. After the study was thoroughly explained to the participants, written informed consent was obtained from both the legal representatives and the individuals. All of the participants met MRI safety criteria.

3.2.2. Clinical assessment

Diagnosis was made according to the DSM-IV criteria using the K-SADS-PL. Clinical diagnostic interviews were performed during the initial hospitalisation by four experienced psychiatrists trained in this interview. Diagnostic consensus was reached in cases where presence or absence of psychiatric diagnoses was in doubt. Psychotic symptoms were assessed using the validated Spanish version of the Positive and Negative Syndrome Scale (PANSS) (Kay et al. 1987). Four psychiatrists participated in the PANSS rating. Single-rater intraclass correlation coefficients for the PANSS ratings were calculated separately for the fixed set of four psychiatrists by comparing each psychiatrist with a gold standard on a randomly selected sample of ten participants. Intraclass correlation coefficients ranged from 0.72 to 0.96. Given that among early onset first-episode psychoses most diagnostic changes occur during the first year after onset of positive symptoms, a 2-year follow-up diagnosis to ensure diagnostic reliability was used (Fraguas et al. 2008). At baseline, 9 participants had schizophrenia, 11 had bipolar I disorder, and 29 had another psychotic disorder. At the 2-year follow-up, 13 participants had schizophrenia,

32 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis

17 had bipolar I disorder and 16 had another psychotic disorder (5 individuals with schizoaffective disorder, 2 with schizophreniform disorder, 1 with major depression with psychotic features, 6 with psychotic disorder NOS and 2 with brief psychotic disorder). Three participants were excluded from the study because they did not fulfil K-SADS-PL criteria for psychosis at the 2-year follow-up, leaving a total sample size of 46 individuals in the psychosis group. The parental socioeconomic status was measured using the Hollingshead–Redlich scale (Hollingshead and Redlich 1958). Assessment of handedness was performed as described in Castro-Fornieles (Castro-Fornieles et al. 2007). The age at onset of psychosis was defined as the age at which the participant experienced positive psychotic symptoms for the first time (delusions or hallucinations of any kind that qualify for a DSM-IV diagnosis). Information about age at onset was assessed using the K-SADS-PL interview by asking the individual and their parents/legal guardians retrospectively about the first appearance of delusions and/or hallucinations. The individual and parents/legal guardians were interviewed separately. If possible, other relatives that were not present at the interview were consulted for confirmation of the gathered information about age at onset. Duration of psychosis was defined as the time difference between age at onset of psychosis and scan acquisition. Duration of treatment was defined as the time between initiation of antipsychotic treatment and scan acquisition.

3.2.3. Medication

At the time of the baseline assessment, all of those in the psychosis group were on antipsychotic medication. In total 80% (n = 39) of the sample (n = 49) were receiving only one antipsychotic, and the remaining 20% (n = 10) were receiving two antipsychotics simultaneously. With the exception of two individuals, the group were receiving a second-generation antipsychotic. Distribution of the antipsychotic treatment was as follows: 51% (n = 25) risperidone, 33% (n = 16) quetiapine, 29% (n = 14) olanzapine, 4% (n = 2) ziprasidone, and 4% (n = 2) haloperidol. The chlorpromazine equivalent dose was calculated from the dose of antipsychotics received (Table 3.1) (Woods 2003). The mean daily antipsychotic dose in chlorpromazine equivalents was 282.4 mg (s.d.= 194.7).

33 Chapter 3

Psychosis Control Statistics group group t-test p (n = 46) (n = 34) Age, years: mean (s.d.) range 15.8 (1.5) 15.1 (1.7) -2.0 0.05 12–18 12–17 Handedness 40/4/0 26/4/1 1.77 0.41 (right/left/ambidexter),A n Level of education, years 8.5 8.5 0 1.00 Parental socioeconomic status,B 2.7 2.9 -0.6 0.55 mean score Age at onset of psychosis,C years: 15.5 (1.5) mean (s.d.) range 10–17 Duration of psychosis,D weeks: 12.9 (10.4) mean (s.d.) range 0–24 Duration of treatment,E weeks: 2.2 (2.0) mean (s.d.) range 0–11 Chlorpromazine equivalents, 282.4 (194.7) mg/day: mean (s.d.) Positive and Negative Syndrome Scale, mean (s.d.) Positive symptoms 25.7 (5.5) Negative symptoms 22.7 (6.7) General psychopathology 48.7 (9.9)

A. Missing data for five participants. B. Parental socioeconomic status was measured using the Hollingshead–Redlich scale. C. Age at onset of psychosis was defined as the age at which positive symptoms appeared for the first time. D. The duration of psychosis was defined as the time between first appearance of positive symptoms and scan acquisition. Missing data for one participant. E. The duration of treatment was defined as the time between start of antipsychotic treatment and scan acquisition.

Table 3.1: Demographic and clinical variables of the adolescent males with early-onset psychosis and the adolescent male controls.

3.2.4. MRI Acquisition

All participants were scanned on the same 1.5 T Philips MRI scanner (Philips Gyroscan; Philips Medical Systems, Best, The Netherlands). Two magnetic resonance sequences were applied to all the participants: a T1-weighted, three-dimensional, gradient echo scan with 1.5mm slice thickness (echo time (TE) = 4.6 ms, repetition time (TR) = 9.3 ms, flip angle 308, field of view (FOV) = 256 mm, and in-plane voxel size 0.986x0.98mm2) and a T2-weighted turbo spin echo scan with 3.5mm slice thickness

34 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis

(turbo factor 15, TE = 120ms, TR= 5809ms, FOV= 256mm, and in-plane voxel size 0.986x0.98mm2). Both T1- and T2-weighted images were used for clinical neurodiagnostic evaluation by an independent neuroradiologist. No participants showed clinically significant brain pathology.

3.2.5. Image Analysis

Intracranial volume assessment The assessment of intracranial volume has been described before (Janssen et al. 2008). Brain-extracted images were obtained using the FMRIB Software Library (FSL)- Brain Extraction Tool (version 2) algorithm (Smith 2002), and then manually edited to remove remaining non-intracranial tissue voxels using in-house Unix software (Desco et al. 2001).

Whole thalamic volume assessment Assessment of thalamic volume was performed by using the FreeSurfer (v4.4.0, http://surfer.nmr.mgh.harvard.edu) package (Fischl et al. 2002). To confirm that the volume findings were independent of the preprocessing method, automated assessment of thalamus volume was also performed using the FSL-FIRST (v4.1.4) package (Patenaude et al. 2011). The current study was focused on the volumes obtained by FreeSurfer. The pattern of detected group differences was similar when using the FSL-FIRST and FreeSurfer methods. Both methods detected a significant volume reduction in the psychosis group and both methods found larger volume differences for the right compared with the left side. This is detailed in Appendix 3.A. Thalamus segmentation using FreeSurfer has been validated through comparison with manual tracing and has been used before in studies assessing thalamic volume (Fischl et al. 2002). Automatic segmentation by FreeSurfer is done by automatic labelling of subcortical tissue classes using an atlas-based Bayesian segmentation procedure. FreeSurfer comes with a participant-independent probabilistic atlas in Talairach space that was pre- computed from a training set of individuals whose brains were manually labelled. FreeSurfer image preprocessing steps included an affine registration with Talairach space, intensity normalisation, skull strip and a high dimensional non-linear volumetric alignment to the probabilistic atlas in Talairach space. Importantly, as normalisation

35 Chapter 3 into Talairach space is an internal component of the FreeSurfer pipeline, the thalamic segmentations are returned in the native space of the individual. FreeSurfer calculated the probability of a class at each voxel location as the probability that the given class appeared at that location in the training set multiplied by the likelihood of getting the participant-specific intensity value from that class. An initial segmentation was generated by assigning each point to the class for which the above probability was greatest. The neighborhood function was then used to recalculate the class probabilities and re-segment the data using the new class probabilities. This procedure was repeated until the result converged (Fischl et al. 2002). FreeSurfer assessment of thalamus volume follows neuroanatomical criteria described in Frazier et al 2007 (Frazier et al. 2005). Briefly, the criteria state that assessment includes all thalamic nuclei except for lateral and medial geniculate bodies. The medial border is the third ventricle, cerebrospinal fluid (CSF) or the cerebral exterior midline. The lateral border is the internal capsule. The anterior border is the interventricular foramen (foramen of Monroe), and posteriorly the thalamus overlaps the midbrain and is bordered by CSF. The inferior border is the hypothalamic fissure, or the hippocampus in the most posterior extent. Superiorly the thalamus assessment is bordered by the transverse cerebral fissure, lateral ventricle, white matter, or in the anterior portion, the caudate. All segmentations were found accurate after visual inspection.

Regional thalamic volume assessment Regional thalamic volume was assessed by generating thalamic surfaces based on Spherical Harmonics Decomposition Point Distribution Models (SPHARM-PDM) (v1.8, www.nitrc.org/projects/spharm-pdm) (Brechbuhler et al. 1995; Styner et al. 2006). This method has been used before in studies investigating regional subcortical volume deficits in people with late-life depression, Alzheimer‟s disease, schizophrenia and individuals at increased genetic risk for schizophrenia (Zhao et al. 2008; Gerardin et al. 2009; Levitt et al. 2009; Ho and Magnotta 2010). The purposes of SPHARM-PDM are: to establish a correspondence between the thalamic surfaces of individuals in an unbiased way, i.e., without using a template (other than for alignment and flipping); to provide a uniform sampling across the surfaces, such that subsequent statistical tests are valid; and to smooth the surfaces by removing high-frequency contributions.

36 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis

Using the method, three parameters are defined by the user: a voxel resolution, SPHARM degree and subdivision level. For the current study an isotropic voxel resolution of 0.5 mm, a SPHARM-PDM degree of 15, and a subdivision level of 10 was chosen. To ensure that the choice for the three parameters was adequate, the effect of modifications of these three parameters on the SPHARM-PDM-based thalamus surface was investigated; this is detailed in Appendix 3.B. The inputs for the method were the thalamus segmentations generated by FreeSurfer. The pipeline consisted of the following steps: first, the three-dimensional thalamus masks were resampled to an isotropic voxel resolution and processed to fill any isolated empty voxels inside the mask. These masks were then converted to boundary surface meshes, and a spherical parameterization was computed from the mesh using an area-preserving, distortion-minimising spherical mapping (Brechbuhler et al. 1995). Second, the surfaces of the masks were represented on the sphere in terms of SPHARM-based shape descriptions (Styner et al. 2006). These shape descriptions characterise the surfaces by sets of coefficients weighting the SPHARM basis functions, i.e., the SPHARM coefficients, up to a certain SPHARM degree. Using the first-order ellipsoid from the SPHARM coefficients, the spherical parameterisations of all participants were aligned to establish correspondence across the surfaces. This alignment was achieved by rotation of the parameterisation, such that the spherical equator, 08 and 908 longitudes coincide with those of the first-order ellipsoid. However, the first-order ellipsoid can be flipped along any of its axes with the same result. Therefore a „flip‟ template (the first surface of the sample, described by SPHARM coefficients) was used to test all possible flips along the first-order ellipsoid axis and select the one whose reconstruction had minimal distance to the „flip‟ template. Third, the SPHARM-based surfaces of all participants were subdivided into an equal number of surface points (SPHARM-PDM) via an icosahedron subdivision level. Although each surface point has a one-to-one correspondence across participants, the surfaces are in the native image space. Therefore, the surfaces must all be aligned to a common space prior to investigating any group differences. To achieve this, a mean thalamus surface computed over all participant‟s first-order ellipsoid-aligned thalamus surfaces was calculated. Thereafter the first-order ellipsoid-aligned SPHARM-PDM-based thalamus surfaces were aligned to the mean surface using Procrustes alignment (translation and rotation to minimise least-square residuals of corresponding three-dimensional points in

37 Chapter 3 the image) (Bookstein 1997). Finally, for the statistical analysis a mean SPHARMPDM- based thalamic surface was estimated from the Procrustes aligned surfaces for each group separately.

3.2.6. Statistics

Statistical analyses examined whether (a) demographic data, (b) intracranial volume, (c) total thalamus volume and (d) distances between the average thalamic surfaces were different between the psychosis group and the control group. Comparison (d) was performed on a point-by-point basis across the entire thalamic surface to address the specificity of regional thalamic volume changes in early-onset psychosis.

Demographic and clinical data To test for group differences in demographic data, the Student‟s t-test was used for the continuous variables and chi-square for discrete categorical variables.

Group-wise analysis of whole thalamus volume To test whether whole thalamic volume reductions were associated with early-onset psychosis, the general linear model with diagnostic group (psychosis group, control group) as the independent variable was used. To test whether whole thalamic volume reductions were associated with schizophrenia or bipolar disorder, the general linear model with diagnostic group (schizophrenia, bipolar disorder) as the independent variable was used. Left and right thalamic volumes were included as repeated measures to evaluate possible interactions between diagnosis and hemisphere. Intracranial volume and age were included as covariates in the analyses as they are known to be related to structural brain measures. Chlorpromazine equivalent dose was examined as a potential covariate but not included because of a lack of significance, i.e., it did not correlate with whole or regional thalamic volume. Effect size is given as Cohen‟s d. For all analyses, the distribution of the dependent measures to ensure they did not deviate from normality was inspected.

Group-wise analysis of regional thalamus volume To test for regional thalamic volume differences, Shape-Analysis-MANCOVA (version 1.0) (http://www.nitrc.org/projects/shape_mancova) was used (Paniagua et al. 2009; Looi et al. 2010; Walterfang et al. 2011) Shape-Analysis-MANCOVA uses a permutation-based approach to compute the local significance values. At each permutation a general linear model-based MANCOVA metric (the Hotelling trace

38 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis coefficient) is computed for every surface (x,y,z) coordinate, which constitutes the local statistic value of the group difference. The distributions sampled via the permutation approach yield the local p-values. This nonparametric approach allows testing for significance even if the surface coordinates are not normally distributed (Paniagua et al. 2009). Intracranial volume and age were included as covariates in the tests. There were 1002 surface points. Owing to the 1002 comparisons, a correction for multiple comparisons was necessary. Shape-Analysis-MANCOVA outputs two results that are corrected for multiple comparisons, false discovery rate (FDR with q = 0.05) corrected results (Genovese et al. 2002) and Bonferroni corrected results. The magnitude in millimetres and direction (i.e., surface inflation or deflation) of the distances before and after correction for multiple comparisons were mapped onto the averaged left and right thalamic surface models of the entire group. To assess whether thalamic measures were related to clinical variables, correlation coefficients were computed.

3.3. Results

3.3.1. Demographic and clinical data

There were no significant group differences in handedness, years of education, and parental socioeconomic status (see Table 3.1). Age was slightly older in the psychosis group. Intracranial volumes were not significantly different in the psychosis group compared with the control group (t = -0.56 (d.f.= 81), p = 0.58). Mean intracranial volumes for each group are shown in Table 3.2.

Mean (s.d.) Psychosis group Control group (n=46) (n=34) Intracranial volumen (cm3) 1517.7 (112.2) 1511.8 (115.0) Thalamus volume (cm3) Left 8.5 (0.8) 8.7 (0.8) Right 8.0 (0.7) 8.3 (0.8)

Table 3.2: Intracranial volume and total left and right thalamus volume of the adolescent males with early-onset psychosis and controls.

39 Chapter 3

3.3.2. Whole thalamic volume

A main effect of diagnosis was observed for overall thalamic volumes after covarying for intracranial volume and age (F (1,76) = 5.68, p = 0.02, d = -0.35). No significant interactions with hemisphere were observed. That is, the psychosis group showed significant thalamic volume reductions in both the left and right hemispheres after correction for age and intracranial volume (left: F (1,76) = 3.78, p = 0.05, d = -0.29, right: F (1,76) = 6.57, p = 0.01, d = -0.41) (Figure 3.1A). Comparing the schizophrenia and bipolar subgroups for overall thalamic volumes while covarying for intracranial volume and age gave no main effect of diagnosis (F (1,26) = 0.64, p = 0.43, d = -0.20) (Figure 3.1B).

Figure 3.1: A) Left and right total thalamus volume of the 46 adolescent males with early-onset psychosis and the 34 adolescent male healthy controls. B) Left and right total thalamus volume of 13 adolescent males with schizophrenia and 17 adolescent males with bipolar I disorder. Note: Values are in cm3, the error bars represent 1 standard deviation. Thalamus volumes were obtained using FreeSurfer software.

3.3.3. Regional thalamic volume

Before correction for multiple comparisons, widespread bilateral regional thalamic volume differences were observed (Figure 3.2A). Only right-sided regional differences survived correction for multiple comparisons (Figure 3.2B and C). Three clusters comprising a total of 24 surface points showed a significant surface deflation in the psychosis group that ranged from 0.39 to 1.05mm after FDR correction. Mean surface deflation over the three clusters was 0.66mm (s.d.= 0.20).

40 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis

Comparing the schizophrenia and bipolar subgroups revealed no significant differences in regional thalamic volumes after correction for multiple comparisons.

Figure 3.2: A) Left and right thalamus regional volumetric differences before correction for multiple comparisons between 34 adolescent male controls and 46 adolescent males with early- onset psychosis. B) and C) Right-sided thalamus regional volumetric differences after correction for multiple comparisons between 34 adolescent male controls and 46 adolescent males with early-onset psychosis. Note: The color bars show the magnitude and direction of the differences (distance in mm) between the average surfaces of each group obtained by SPHARM- PDM. A positive distance means that the average surface of the individuals in the early-onset psychosis group represents contraction with respect to the average surface of the controls (i.e., surface deflation of those in psychosis group) and vice versa.

41 Chapter 3

3.3.4. Correlation with clinical variables

Within the psychosis group, both total and regional thalamic measures were not significantly related to age at onset of psychosis, duration of psychosis, duration of treatment and PANSS scores.

3.4. Discussion

The main findings of this study were, first, that male adolescents with early-onset first-episode psychosis showed a bilateral whole thalamus volumetric deficit. Second, the psychosis group demonstrated statistically significant right-sided regional thalamic volume differences in areas corresponding to the anterior mediodorsal and pulvinar nuclei when compared with the control group.

Whole thalamus volume These findings are congruent with previous MRI reports of smaller total thalamus volume in adolescents with early-onset schizophrenia.(Dasari et al. 1999; Kumra et al. 2000; Frazier et al. 2008; Ellison-Wright and Bullmore 2010). For early-onset bipolar disorder prior findings are inconsistent (Dasari et al. 1999; Chang et al. 2005; Frazier et al. 2008). The present study contrasts with previous early-onset bipolar disorder reports in that those studies examined mixed diagnostic samples including individuals with nonpsychotic bipolar disorder and individuals with a longer duration of illness and extensive medicinal treatment, including lithium. These differences, in particular the prolonged use of medication, may explain the discrepancy in results because exposure to typical and atypical antipsychotics was shown to have rapid opposite effects on thalamic volume in adults with first episode psychosis and schizophrenia and cerebral biochemical changes were found after treatment with lithium and risperidone in early- onset bipolar disorder (Dazzan et al. 2005; Terry et al. 2009). These results indicate that prolonged medication use can change cerebral structure. The current results denote that volumetric thalamic deficits are present shortly after symptom onset in minimally treated adolescents with psychosis who go on to develop schizophrenia and bipolar I disorder. Although there was no interaction between diagnosis and hemisphere, the effect size of the left side was lower compared with the right side. The effect of limited sample sizes on detection of subtle thalamic volume differences has been noted and may also be

42 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis a plausible explanation for the difference in effect sizes (Byne et al. 2008). A role for brain asymmetry in the etiology of psychosis has been suggested but larger left than right thalamus volumes and the reverse have been reported, making it difficult to conclude whether lateralised thalamus volume deficits have any role in this pathology (Flaum et al. 1995; Gur et al. 1998; Crow 2008).

Regional thalamus volume The surface-based analysis showed statistically significant regional volume differences in the anterior mediodorsal and pulvinar regions between those in the psychosis group and the control group. The psychosis group showed surface deflation relative to the control group in these areas, which indicates a smaller volume. Structural abnormalities in these regions, including reduced volume, have been reported in post-mortem and in in vivo MRI studies in medicated and medication-naive adults with first episodes of psychosis and schizophrenia (Harms et al. 2007; Coscia et al. 2009). As can be seen in Figure 3.2A, the statistically uncorrected maps indicated widespread bilateral differences between the psychosis group and the control group, including anterior and pulvinar regions. These results must, however, be interpreted with caution as they are not appropriately corrected for multiple comparisons.

Anterior mediodorsal thalamic region and pulvinar The thalamus is an important hub in the communication network between distinct associative cortical areas. With regard to psychosis, the anterior mediodorsal and pulvinar regions combined occupy approximately one-third of the total thalamus volume and both anterior and posterior thalamic regions have abundant reciprocal connections with the prefrontal, temporal, parieto-occipital and the entorhinal cortex (Romanski et al. 1997; Behrens et al. 2003). Functional and structural prefrontal impairments are hallmark deficits of psychosis. These deficits may be associated with the regional structural thalamic impairments seen in the current study. In an explorative analysis a significantly smaller prefrontal cortex volume in individuals with early-onset psychosis when compared with controls was found, although prefrontal cortex volume and thalamus volume were not correlated (results not shown). A relationship between prefrontal structural impairment and thalamic impairment thus remains speculative but evidence in favour of such a relationship is coming from

43 Chapter 3 studies measuring the integrity of prefrontothalamic white matter connections using diffusion tensor imaging. These studies show that smaller volume of prefrontal grey matter structures is associated with deficits of the adjoining corticothalamic white matter tracts in adolescents with first-episode schizophrenia (Douaud et al. 2007). Furthermore, changes in prefrontal dopaminergic neurotransmission are marked in individuals with psychosis and recent studies in humans and non-human primates have shown that the anterior mediodorsal area is a thalamic region with a particularly high level of dopaminergic innervation (Garcia-Cabezas et al. 2007). Dopaminergic projections to the thalamus and prefrontal cortex have the same origin and the thalamic and prefrontal regions that receive these projections are highly reciprocally connected. Speculatively, alterations in prefrontal dopaminergic neurotransmission may thus influence dopaminergic receptivity in the thalamic anterior mediodorsal region that could lead to impairment of corticothalamic circuits in individuals with psychosis (Alelu-Paz and Gimenez-Amaya 2008). Thalamic structural abnormalities in adolescents with early onset psychosis, together with findings of smaller thalamus volume in young individuals with an increased genetic risk for psychosis, suggest that thalamic volumetric deficits are a vulnerability marker for the disease (Ettinger et al. 2007). This interpretation is supported by longitudinal studies that have reported baseline thalamic volume deficits to be non-progressive (James et al. 2004), although others have not replicated this (Rapoport et al. 1997). In the current study, the reported effect sizes for whole thalamic volume reduction are small to moderate (Cohen‟s d -0.3 to -0.4). These effect sizes are comparable with previous meta-analytically estimated effect sizes for reduction of thalamus and hippocampal volume but considerably lower than effect sizes of increased lateral ventricle volume (d = 0.7) in adults with first-episode and chronic schizophrenia (Wright et al. 2000; Adriano et al. 2010). The cross sectional design of the current study impedes an interpretation as to whether thalamic volume deficits represent increased sensitivity for psychosis or whether the deficits are a consequence of psychosis or related variables. Regarding the effect of patient subgroups, only two prior studies that compared volume of subcortical brain structures between youths with early-onset bipolar disorder, schizophrenia and healthy controls (Frazier et al. 2008; Janssen et al. 2008) are known. Frazier et al 2008 (Frazier et al. 2008) reported a smaller right-sided thalamic volume in

44 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis the schizophrenia group whereas the other study did not find a group difference in thalamic volume (Janssen et al. 2008). In the current study no group differences were found between the bipolar and schizophrenia subgroups. Recent and older meta-analyses indicate a modest effect size (Cohen‟s d = 0.4) for thalamic volume reduction in schizophrenia (Konick and Friedman 2001; Adriano et al. 2010). Therefore the reduced sample size when comparing patients only is too small to detect differences of d = 0.4 with high statistical power.

Limitations There are several limitations to this study. First, any manual tracings of the thalamus were used as a gold standard in order to estimate the accuracy of the automated segmentations. However, volume differences between groups were confirmed by two different automated methods, thus strengthening the reliability of the findings. Second, the patient subgroups comparison suffers from insufficient statistical power because of the small sample sizes. Third, post-mortem studies are necessary in relating markers to the underlying disease process and the ultimate proof can only be provided by pathological validation of these markers.

Acknowledgements

The authors declare no competing financial interests. The authors would like to thank all study participants and their families. This study is supported by the Instituto de Salud Carlos III, the Juan de la Cierva Program of the Spanish Ministry of Science and Innovation, the Fundación Alicia Koplowitz, and Caja Navarra. Part of the computations of this work was done at the „High Capacity Cluster for Physical Techniques‟ of the Universidad Complutense de Madrid (UCM), funded in part by the European Union (Fondo Europeo de Desarrollo Regional program) and UCM.

45 Chapter 3

Related appendices

Appendix 3.A. Thalamic volume assessment and results using FSL-FIRST.

Automated assessment of total thalamus volume was performed twice; once using FreeSurfer and once using the FSL-FIRST (v4.1.4) package. Thalamus segmentation using FreeSurfer and FSL-FIRST has been validated through comparison with manual tracing. Automatic segmentation by FreeSurfer is done by automatic labelling of subcortical tissue classes using an atlas based Bayesian segmentation procedure. FreeSurfer comes with a subject-independent probabilistic atlas in Talairach space that was pre-computed from a training set of individuals whose brains were manually labelled. FreeSurfer image preprocessing steps included an affine registration with Talairach space, intensity normalization, skull strip and a high dimensional non-linear volumetric alignment to the probabilistic atlas in Talairach space. FreeSurfer calculated the probability of a class at each voxel location as the probability that the given class appeared at that location in the training set multiplied by the likelihood of getting the subject-specific intensity value from that class. An initial segmentation was generated by assigning each point to the class for which the above probability was greatest. The neighbourhood function was then used to recalculate the class probabilities and resegment the data using the new class probabilities. This procedure was repeated until the result converged (Patenaude et al. 2011). The FSL-FIRST method is a probabilistic adaptation of the active appearance model (Fischl et al. 2002). The method is informed by the shape and intensity variations of a structure from a training set for the purpose of automatically segmenting the structure. A multivariate Gaussian model of vertex location and intensity variation is used, and is based on having point correspondence across individuals (same number and labelling of vertices across individuals). The necessary correspondence is imposed during the parameterization of the labelled images with a deformable model. The model is fit to new images by maximising the posterior probability of shape given the observed intensities. All segmentations were found to be accurate after visual inspection.

Whole thalamus volume analysis from FSL-FIRST A main effect of diagnosis was observed for overall thalamic volumes after covarying for intracranial volume and age (F (1.76) = 5.32, p= 0.02, d = -0.35). No significant interactions with hemisphere were observed. That is, after correction for intracranial volume and age, individuals in the

46 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis psychosis group showed a trend-level thalamic volume reduction in the left and a significant reduction in the right hemisphere (left: F (1,76) = 3.32, p = 0.07, d =-0.29, right: F (1,76) = 7.43, p = 0.008, d = -0.46) (See Figure 3.A.1 and Figure 3.A.2).

Figure 3.A.2: A regression plot showing total thalamus volume from FreeSurfer vs. FSL-FIRST (FSL) data-sets, along with the correlation coefficient and regression equation relating the two sets of volumes.

47 Chapter 3

Appendix 3.B. Determination of an adequate combination of parameters for generating spherical harmonics (SPHARM)-based thalamic surfaces

The goal was to find an adequate combination of the three parameters that are chosen by the user (voxel resolution, SPHARM degree and subdivision level) for calculating the SPHARM-based thalamus surfaces. The SPHARM degree determines the number of spherical basis functions that are needed to express the initial (not SPHARM-based) boundary surface meshes. If the SPHARM degree is increased, a higher number of basis functions will be used and the surface representation will be sharper. If the SPHARM degree decreases, a smoother surface representation will be obtained. The subdivision level represents the number of points of the spherical grid for the surface spherical mapping. If the subdivision level is increased, the SPHARM-based surface will have a higher number of points allowing for a better spatial representation of the initial boundary surface. In case of the thalamus, which has a smooth surface, a high subdivision level is not mandatory. Previous studies using SPHARM to assess the shape of subcortical structures have suggested a SPHARM degree ranging from 12 to 15 and a subdivision level ranging from 10 to 20 as adequate(Styner et al. 2004; Styner et al. 2005). To find an adequate combination of the three parameters for the current study, SPHARM-based surfaces were created using different parameter combinations within the proposed ranges. FreeSurfer assessments of the thalamus were resampled to three voxel resolutions (0.5, 0.75 and 1 mm). Five SPHARM degrees (7, 12, 13, 14, 15) and three subdivision levels (10, 15, 20) were combined with the three-voxel resolution parameters. The SPHARM-based surfaces were generated in native space for each possible combination of the voxel resolution, SPHARM degree and subdivision-level parameters. For every parameter combination, the absolute distance in millimeters between the SPHARM-based surface and the initial (not SPHARM) surface was computed for each surface element. The distance was calculated as the length of the normal vector at each surface element from the SPHARM-based surface until its corresponding intersection point with the initial (not SPHARM) surface. A shorter distance reflects a good approximation of the initial (not SPHARM) surface by the SPHARM-based surface. For each parameter combination, the distance was averaged over all surface elements and over all 80 participants. This mean distance was then outlined against all combinations of the parameters. As can be seen in Figure 3.B.1, changing the voxel resolution and the subdivision level did not strongly affect the mean

48 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis distance. Based on these findings, a voxel resolution of 0.5x0.5x0.5mm3, SPHARM degree of 15, and a subdivision level of 10 was found to be an adequate combination of parameters for generating the SPHARM-based surfaces.

49 Chapter 3

Related references

Adriano, F., I. Spoletini, et al. (2010). "Updated meta-analyses reveal thalamus volume reduction in patients with first-episode and chronic schizophrenia." Schizophrenia Research 123(1): 1-14. Alelu-Paz, R. and J. M. Gimenez-Amaya (2008). "The mediodorsal thalamic nucleus and schizophrenia." Journal of Psychiatry & Neuroscience 33(6): 489-498. APA (1994). "Diagnostic and statistical manual of mental disorders." Washington, DC. Behrens, T. E., H. Johansen-Berg, et al. (2003). "Non-invasive mapping of connections between human thalamus and cortex using diffusion imaging." Nature Neuroscience 6(7): 750- 757. Bookstein, F. (1997). "Shape and the information in medical images: a decade of the morphometric synthesis." Computer Vision and Image Understanding 66: 97-118. Brechbuhler, C., G. Gerig, et al. (1995). "Parametrization of closed surfaces for 3-D shape description." Computer Vision, Graphics and Image Processing 61: 154-170. Byne, W., E. A. Hazlett, et al. (2009). "The thalamus and schizophrenia: current status of research." Acta Neuropathologica 117(4): 347-368. Castro-Fornieles, J., M. Parellada, et al. (2007). "The child and adolescent first-episode psychosis study (CAFEPS): design and baseline results." Schizophrenia Research 91(1- 3): 226-237. Coscia, D. M., K. L. Narr, et al. (2009). "Volumetric and shape analysis of the thalamus in first- episode schizophrenia." Human Brain Mapping 30(4): 1236-1245. Crow, T. J. (2008). "The 'big bang' theory of the origin of psychosis and the faculty of language." Schizophrenia Research 102(1-3): 31-52. Chang, K., A. Karchemskiy, et al. (2005). "Reduced amygdalar gray matter volume in familial pediatric bipolar disorder." Journal of the American Academy of Child & Adolescent Psychiatry 44(6): 565-573. Dasari, M., L. Friedman, et al. (1999). "A magnetic resonance imaging study of thalamic area in adolescent patients with either schizophrenia or bipolar disorder as compared to healthy controls." Psychiatry Research 91(3): 155-162. Dazzan, P., K. D. Morgan, et al. (2005). "Different effects of typical and atypical antipsychotics on grey matter in first episode psychosis: the AESOP study." Neuropsychopharmacology 30(4): 765-774. Desco, M., J. Pascau, et al. (2001). "Multimodality Image Quantification using Talairach Grid." Proc. SPIE Medical Imaging. M. Sonka and K. M. Hanson. 4322: 1385-1392. Douaud, G., S. Smith, et al. (2007). "Anatomically related grey and white matter abnormalities in adolescent-onset schizophrenia." Brain 130(Pt 9): 2375-2386. Ellison-Wright, I. and E. Bullmore (2010). "Anatomy of bipolar disorder and schizophrenia: a meta-analysis." Schizophrenia Research 117(1): 1-12. Ettinger, U., M. Picchioni, et al. (2007). "Magnetic resonance imaging of the thalamus and adhesio interthalamica in twins with schizophrenia." Archives of General Psychiatry 64(4): 401-409. Fischl, B., D. H. Salat, et al. (2002). "Whole brain segmentation: automated labeling of neuroanatomical structures in the human brain." Neuron 33(3): 341-355.

50 Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis

Flaum, M., V. W. Swayze, 2nd, et al. (1995). "Effects of diagnosis, laterality, and gender on brain morphology in schizophrenia." American Journal of Psychiatry 152(5): 704-714. Fraguas, D., M. J. de Castro, et al. (2008). "Does diagnostic classification of early-onset psychosis change over follow-up?" Child Psychiatry & Human Development 39(2): 137-145. Frazier, J. A., S. Chiu, et al. (2005). "Structural brain magnetic resonance imaging of limbic and thalamic volumes in pediatric bipolar disorder." American Journal of Psychiatry 162(7): 1256-1265. Frazier, J. A., S. M. Hodge, et al. (2008). "Diagnostic and sex effects on limbic volumes in early-onset bipolar disorder and schizophrenia." Schizophrenia Bulletin 34(1): 37-46. Garcia-Cabezas, M. A., B. Rico, et al. (2007). "Distribution of the dopamine innervation in the macaque and human thalamus." Neuroimage 34(3): 965-984. Genovese, C. R., N. A. Lazar, et al. (2002). "Thresholding of statistical maps in functional neuroimaging using the false discovery rate." Neuroimage 15(4): 870-878. Gerardin, E., G. Chetelat, et al. (2009). "Multidimensional classification of hippocampal shape features discriminates Alzheimer's disease and mild cognitive impairment from normal aging." Neuroimage 47(4): 1476-1486. Gur, R. E., V. Maany, et al. (1998). "Subcortical MRI volumes in neuroleptic-naive and treated patients with schizophrenia." American Journal of Psychiatry 155(12): 1711-1717. Harms, M. P., L. Wang, et al. (2007). "Thalamic shape abnormalities in individuals with schizophrenia and their nonpsychotic siblings." Journal of Neuroscience 27(50): 13835- 13842. Ho, B. C. and V. Magnotta (2010). "Hippocampal volume deficits and shape deformities in young biological relatives of schizophrenia probands." Neuroimage 49(4): 3385-3393. Hollingshead, A. and F. Redlich (1958). "Social Class and Mental Illness: A Community Study." James, A. C., S. James, et al. (2004). "Cerebellar, prefrontal cortex, and thalamic volumes over two time points in adolescent-onset schizophrenia." American Journal of Psychiatry 161(6): 1023-1029. Janssen, J., S. Reig, et al. (2008). "Regional gray matter volume deficits in adolescents with first-episode psychosis." Journal of the American Academy of Child & Adolescent Psychiatry 47(11): 1311-1320. Jones, E. G. (1997). "Cortical development and thalamic pathology in schizophrenia." Schizophrenia Bulletin 23(3): 483-501. Kaufman, J., B. Birmaher, et al. (1997). "Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version (K-SADS-PL): initial reliability and validity data." Journal of the American Academy of Child & Adolescent Psychiatry 36(7): 980-988. Kay, S. R., A. Fiszbein, et al. (1987). "The positive and negative syndrome scale (PANSS) for schizophrenia." Schizophrenia Bulletin 13(2): 261-276. Konick, L. C. and L. Friedman (2001). "Meta-analysis of thalamic size in schizophrenia." Biological Psychiatry 49(1): 28-38. Kumra, S., J. N. Giedd, et al. (2000). "Childhood-onset psychotic disorders: magnetic resonance imaging of volumetric differences in brain structure." American Journal of Psychiatry 157(9): 1467-1474.

51 Chapter 3

Levitt, J. J., M. Styner, et al. (2009). "Shape abnormalities of caudate nucleus in schizotypal personality disorder." Schizophrenia Research 110(1-3): 127-139. Looi, J. C., M. Walterfang, et al. (2010). "Shape analysis of the neostriatum in frontotemporal lobar degeneration, Alzheimer's disease, and controls." Neuroimage 51(3): 970-986. Paniagua, B., M. Styner, et al. (2009). "Local Shape Analysis using MANCOVA." Insight Journal 694. Patenaude, B., S. M. Smith, et al. (2011). "A Bayesian model of shape and appearance for subcortical brain segmentation." Neuroimage 56(3): 907-922. Rapoport, J. L., J. Giedd, et al. (1997). "Childhood-onset schizophrenia. Progressive ventricular change during adolescence." Archives of General Psychiatry 54(10): 897- 903. Romanski, L. M., M. Giguere, et al. (1997). "Topographic organization of medial pulvinar connections with the prefrontal cortex in the rhesus monkey." Journal of Comparative Neurology 379(3): 313-332. Smith, S. M. (2002). "Fast robust automated brain extraction." Human Brain Mapping 17(3): 143-155. Styner, M., J. A. Lieberman, et al. (2004). "Boundary and medial shape analysis of the hippocampus in schizophrenia." Medical Image Analysis 8(3): 197-203. Styner, M., J. A. Lieberman, et al. (2005). "Morphometric analysis of lateral ventricles in schizophrenia and healthy controls regarding genetic and disease-specific factors." Proceedings of the National Academy of Sciences 102(13): 4872-4877. Styner, M., I. Oguz, et al. (2006). "Framework for the Statistical Shape Analysis of Brain Structures using SPHARM-PDM." Insight Journal (1071): 242-250. Terry, J., M. Lopez-Larson, et al. (2009). "Magnetic resonance imaging studies in early onset bipolar disorder: an updated review." Child & Adolescent Psychiatric Clinics of North America18(2): 421-439, ix-x. Walterfang, M., J. C. Looi, et al. (2011). "Shape alterations in the striatum in chorea- acanthocytosis." Psychiatry Research 192(1): 29-36. Wechsler, D. (1974). "Manual for the Wechsler Intelligence Scale for Children." The Psychological Corporation. Wechsler, D. (1997). "Wecshler Adult Intelligence Scale." The Psychological Corporation. Woods, S. W. (2003). "Chlorpromazine equivalent doses for the newer atypical antipsychotics." Journal of Clinical Psychiatry 64(6): 663-667. Wright, I. C., S. Rabe-Hesketh, et al. (2000). "Meta-analysis of regional brain volumes in schizophrenia." American Journal of Psychiatry 157(1): 16-25. Zhao, Z., W. D. Taylor, et al. (2008). "Hippocampus shape analysis and late-life depression." PLoS One 3(3): e1837.

52 The human cortex flattens during adolescence

Chapter 4 The human cortex flattens during adolescence

Yasser Alemán-Gómez1, Joost Janssen1, Hugo Schnack, Evan Balaban, Laura Pina- Camacho, Fidel Alfaro-Almagro, Josefina Castro-Fornieles, Soraya Otero, Immaculada Baeza, Dolores Moreno, Nuria Bargalló, Mara Parellada, Celso Arango, and Manuel Desco

Journal of Neuroscience. Sep 18; 33(38):15004-10. doi: 10.1523/Jneurosci.1459-13.2013. 1These authors contributed equally to this study.

The human cortex flattens during adolescence

Abstract

The human cerebral cortex appears to shrink during adolescence. To delineate the dynamic morphological changes involved in this process, 52 healthy male and female adolescents (11-17 years) were neuroimaged twice using Magnetic Resonance Imaging, approximately two years apart. Using a novel morphometric analysis procedure combining the FreeSurfer and BrainVISA image software suites, we quantified global and lobar change in cortical thickness, outer surface area, the gyrification index, the average Euclidean distance between opposing sides of the white matter surface (gyral white matter thickness), the convex („exposed‟) part of the outer cortical surface (hull surface area), sulcal length, depth and width. Obtained results suggest that the cortical surface flattens during adolescence. Flattening was strongest in the frontal and occipital cortex, where significant sulcal widening and decreased sulcal depth co-occurred. Globally, sulcal widening was associated with cortical thinning and, for the frontal cortex, with loss of surface area. For the other cortical lobes, thinning was related to gyral white matter expansion. The overall flattening of the macrostructural three- dimensional architecture of the human cortex during adolescence thus involves changes in grey matter and effects of the maturation of white matter.

55 Chapter 4

4.1. Introduction

Adolescence is a period of important cortical brain changes for which longitudinal MRI studies are ideally suited. Recent findings suggest that during adolescence in both sexes, the cortex globally contracts due to cortical thinning (Shaw et al. 2008; Blakemore 2012; Brown et al. 2012; van Soelen et al. 2012), which may be highly heritable (van Soelen et al. 2012). Annual reductions are found to be higher than 1% across most parts of the cortex (van Soelen et al. 2012; Tamnes et al. 2013) and may follow a posterior-anterior pattern with medial and dorsolateral prefrontal areas the last to show decline (Gogtay et al. 2004; Tamnes et al. 2013). During adolescence, development of gyral and sulcal surface area shows a more subtle decline compared to cortical thickness (Raznahan et al. 2011; Brown et al. 2012). Subcortical white matter volume continues to expand linearly in most but not all subcortical regions from childhood to early adulthood (Lenroot et al. 2007; Tamnes et al. 2010; Brouwer et al. 2012; Brown et al. 2012). The direct relationship among these processes is not well understood, nor is it known whether such a relationship is invariant over different cortical regions; better knowledge about these issues could provide important insights into the dynamics of cortical development during adolescence. This study used novel surface-based morphometric methods that yielded two improvements with respect to prior surface-based studies focusing on cortical thickness and surface area during brain development. First, the present analysis incorporates a new and detailed set of sulcal measurements in addition to gyral measures of thickness and surface area. Cortical surface area varies between individuals because of differences in either „exposed‟ (gyral) or „hidden‟ (sulcal) surface area or both. Hidden sulcal surface area makes up the largest part of the human cortical surface (Van Essen and Drury 1997) and varies between individuals because of changes in sulcal depth, length, and width. Second, cortical thinning during adolescence may partly be an artifact of the maturation of white matter tracts that cause voxels at the interface between white and gray matter to change their classification (Sowell et al. 2004; Shaw et al. 2008; Paus 2010). The present analysis simultaneously examines the changes over time in cortical surface area, thickness, the gyrification index, hull surface area, gyral white matter thickness, sulcal depth, length, and width in order to gain better insight into both the overall dynamics of cortical development and to examine differences between different

56 The human cortex flattens during adolescence cortical regions during adolescence. The direct dependency between gyral white matter thickness and cortical thickness was also examined to clarify to what degree cortical changes might be attributable to white matter maturation. Longitudinal (two-year) changes in these parameters were examined in 52 typically-developing male and female adolescents (all younger than 18 years).

4.2. Materials and methods

MRI data were collected as part of a two-year longitudinal multi-centre study of First-Episode Psychosis with onset in adolescence which has been comprehensively described elsewhere (Castro-Fornieles et al. 2007). Healthy controls were recruited from publicly-funded schools in the community. Subjects were offered a coupon to buy school supplies in compensation for their participation and a trained psychologist conducted a preliminary telephone screening to check for inclusion criteria. Those who passed the initial screening were interviewed with their relatives at the clinical centers by experienced child and adolescent psychiatrists. The inclusion criteria were an age between 7 and 17 years at the time of first evaluation, no current or previous psychiatric disorder as measured by the Kiddie Schedule for Affective Disorders and Schizophrenia, Present and Lifetime version (K- SADS-PL, a semi-structured diagnostic interview designed to assess current and past psychopathologic conditions (Kaufman et al. 1997), and no neurological disorders, head trauma or mental retardation based on DSM-IV criteria (APA 1994). The study was approved by the institutional review boards of participating clinical centers. After the study was thoroughly explained to the participants, written informed consent was obtained from both the legal representatives and the individuals (if older than 16 years of age). All participants met MRI safety criteria. A total of 98 subjects were initially included of which a subsample of 70 subjects completed baseline and longitudinal (two-year) imaging. Image quality was checked before image processing using two tools. 1) The “Check sample homogeneity” tool in the SPM-VBM8 toolbox (v.r435, http://dbm.neuro.uni-jena.de/vbm/check-sample- homogeneity/). This tool calculates the standard deviation by the sum of the squared distance of each image from the sample mean. Images that were more than two standard deviations from the mean were checked visually and excluded if deemed of insufficient quality. 2) The Freesurfer QA tool (v5.1, http://surfer.nmr.mgh.harvard.edu/fswiki/- QATools). This tool generates snapshots of anatomically labelled surfaces which were

57 Chapter 4 checked for major topological defects and label accuracy. Images were excluded if deemed of insufficient quality. 18 (mean age (SD) 15.1 (1.6) years, 5 females) of the original 70 subjects had insufficient image quality and were excluded from the study, leaving a sample of 52 subjects (20 females, see Table 4.1).

Subjects sample (n=52) Age at baseline (Years) 15.4 ± 1.5 Age at follow-up (Years) 17.5 ± 1.6 Interscan interval (Years) 2.1 ± 0.3 Age range 11 -- 17 Sex, n (% male) 32 (61.5) Handedness, n (% right) 44 (84.6) Parental education (years) 14.1 ± 4.1 Parental socioeconomic status 6/16/15/3/12 (1/2/3/4/5), n Estimated IQ 108 ± 17.6 IQ Range 73 -- 141 Ethiticity (Hispanic/Caucasian/Other) 3/48/1

Table 4.1: Sociodemographics at baseline of the sample.

4.2.1. Clinical and functional assessment

The diagnostic and clinical assessments and the functional assessments (including IQ) were performed at the corresponding site by trained psychiatrists or neuropsychologists, respectively, at baseline and followup. The rater was the same for each subject at baseline and at the 2-year follow-up assessment visit. Each neuropsychologist had been previously trained in the use of the IQ scales before starting assessment and had to reach good reliability with other previously trained neuropsychologists (an intraclass correlation coefficient higher than 0.80). Diagnosis was established according to DSM-IV criteria, using the Spanish K-SADS-PL. Parents and healthy controls were interviewed separately by psychiatrists trained in the use of this instrument. Diagnostic consensus was achieved when the presence or absence of a psychiatric disorder was in doubt. The Pervasive Developmental Disorder exclusion diagnoses were made following clinical and DSM-IV criteria. The vocabulary, Information and Block Design subtests of the Wechsler Intelligence Scale for Children (WISC-R) or the Wechsler Adult Intelligence Scale

58 The human cortex flattens during adolescence

(WAIS-III) were used to estimate the IQ of those younger than 16 years or 16 years and older, respectively (Ringe et al. 2002) and is reported to show good correspondence with full-scale IQ (Satler 2001). Handedness was assessed by means of item five of the Neurological Evaluation Scale (NES) (Buchanan and Heinrichs 1989).

4.2.2. MRI Acquisition

All subjects had their two scans acquired using the same scanner and the acquisition parameters are shown in the next Table (Table 4.2). Data were collected at each centre and processed at one site only. Two magnetic resonance sequences were acquired for all the participants: a 3D T1-weighted sagittal MPRAGE and a 3D T2- weighted axial turbo spin echo.

Scanner N(Males) Age at Voxel size Field of TR TE Flip baseline (mm3) View (mm2) (ms) (ms) Angle (0) Philips 26(16) 15.3 ± 1.4 T1 1.0x1.0x1.5 256x256 15.3 4.6 30 Gyroscan T2 1.0x1.0x3.5 256x256 5.8 120 90 ACS 1.5T Siemens 14(8) 15.9 ± 0.9 T1 1.0x1.0x1.5 256x256 18.1 2.39 20 Symphony T2 0.98x0.98x3.5 256x180 5.8 120 150 1.5T GE 4(2) 13.8 ± 2.1 T1 0.98x0.98x1.5 250x250 12.1 5.2 20 Genesis T2 0.98x0.98x3.5 250x250 5.8 126 90 Signa 1.5T GE 6(5) 16.7 ± 0.5 T1 0.98x0.98x1.5 250x250 10.9 4.6 20 Genesis T2 0.98x0.98x3.5 250x250 5.8 126 90 Signa 1.5T Siemens 2(1) 15.9 ± 0.9 T1 0.98x0.98x2.0 250x250 20.0 5.04 15 Symphony T2 1.0x1.0x3.5 256x192 5.8 116 150 1.5T

Table 4.2: Acquisition parameters as well as number of subjects for each scanner.

Both T1- and T2-weighted images were used for clinical neurodiagnostic evaluation by an independent neuroradiologist. No participants showed clinically significant brain pathology.

59 Chapter 4

4.2.3. Image analysis

Lobar cortical thickness, surface area, gyrification index, and gyral WM thickness The FreeSurfer longitudinal stream (v5.1) was used to generate accurate unbiased baseline and follow-up white and pial surfaces, their voxel-based representations, the „ribbon‟ image (Reuter and Fischl 2011; Reuter et al. 2012) and a lobar cortical parcellation (Dale et al. 1999; Fischl et al. 1999; Desikan et al. 2006; Reuter and Fischl 2011; Reuter et al. 2012). In the FreeSurfer cross-sectional analysis each time point would be processed independently for each subject. These processes involve solving many complex nonlinear optimization problems that are typically calculated using iterative methods. Such methods need starting conditions that may introduce biases in the final results. The FreeSurfer longitudinal stream is designed to minimize bias with respect to any time point in a subject. The longitudinal analysis uses results from the cross-sectional analysis and consists of two main steps: i) creation of a template for each subject using all time points to build an average subject anatomy and ii) analysis of each time point using information from the template and the individual cross-sectional runs to initialize several of the segmentation algorithms. This procedure of using the repeated measures as common information from the subject to initialize the processing in each time point can reduce variability compared to independent processing, as has been shown recently (Reuter et al. 2012). Lobar cortical thickness (CT) and lobar pial surface area (SA) were calculated from the white and gray matter surface (see Figure 4.1). FreeSurfer measurements of CT and SA have been validated via histological and manual measurements and have demonstrated to show good test-retest reliability across scanner manufacturers (Rosas et al. 2002; Kuperberg et al. 2003; Han et al. 2006).

60 The human cortex flattens during adolescence

Figure 4.1: Schematic representation of a gyrus and sulcus representing the different cortical morphometrics used in the current study.

The lobar gyrification index (GI) was calculated to assess the degree of lobar gyrification (see Figure 4.2) (Zilles et al. 1988). In order to generate a smooth envelope that wrapped around the hemisphere but did not encroach into the sulci, a morphological isotropic closing of 6 mm was applied recursively to the cortical parcelation („aparc+aseg‟) image to ensure boundary smoothness. Thereafter, an unlabeled hemispheric envelope was created using the marching cubes algorithm. Secondly, the hemispheric envelope was parcellated into brain lobes following the approach described by Su et al (Su et al. 2013). This regional parcellation allows for computing the hull surface area (HS) for each lobe. The GI is defined as the lobar cortical surface area divided by the lobar hemispheric hull surface area and represents the amount of sulcal surface area.

61 Chapter 4

Figure 4.2: Parcellated hemispheric cortical surface overlaid with parcellated wire-frame representation of the cortical surface hull.

Lobar gyral WM thickness (WT) was estimated using a medial gyral WM surface that was generated using the FreeSurfer WM segmentation derived from the „ribbon‟ image. The medial gyral WM surface is the surface that transverses the gyral WM space, parallel to the gyral GM/WM borders and covers the entire gyral “depth” from crest to base. Lobar WT is calculated as the Euclidean distance between two points residing on opposing sides of the FreeSurfer white surface in the direction normal to the medial gyral WM surface averaged over all gyri pertaining to a lobe (see Figure 4.1) (Kochunov et al. 2009; Kochunov et al. 2012).

Lobar sulcal depth, length and width The „ribbon‟ images of each subject were imported into BrainVISA ‟s (v4.2.1) Morphologist 2012 pipeline (see Figure 4.3). Using default settings, the „ribbon‟ image was used to generate a GM/CSF mask from which the cortical sulci were then automatically segmented throughout the cortex, with the cortical sulci corresponding to the crevasse bottoms of the “landscape”, the altitude of which is defined by image intensity. This definition provides a stable and robust median sulcal surface definition that is not affected by variations in the cortical thickness or width, or by the GM/WM contrast (Mangin et al. 2004; Jouvent et al. 2011). The median sulcal surface spans the

62 The human cortex flattens during adolescence entire space contained in a sulcus, from the fundus to its intersection with the hull. Median sulcal surfaces were automatically labeled (Perrot et al. 2011). Thereafter the median sulcal surfaces were re-labeled into lobes by assigning each labeled sulcus to a lobe based on an a-priori definition. Lobar sulcal depth (SD) is defined as the geodesic distance between the fundus and the hull averaged over all points along all median sulcal surfaces pertaining to a lobe (see Figure 4.1 and Appendix 4.A) (Jouvent et al. 2011; Kochunov et al. 2012). Lobar sulcal length (SL) is measured on the hull and is defined as the distance of the median sulcal surface intersecting the hull, summed over all median sulcal surfaces pertaining to a lobe (see Figure 4.1 and Appendix 4.A) Lobar sulcal width (SW) is defined as the distance between each gyral bank averaged over all points along all median sulcal surfaces pertaining to a lobe (see Figure 4.1 and Appendix 4.A) (Jouvent et al. 2011; Liu et al. 2011; Kochunov et al. 2012).

Figure 4.3: Schematic representation of the image processing. FreeSurfer (v5.1) and BrainVISA (v4.2.1) software were combined.

63 Chapter 4

All of the measurements were either summed (for SA, HS, and SL) or averaged (for CT, GI, WT, SD, and SW) across hemispheres. All of the measurements were done in the native space of the subject‟s images. For each subject, all of the image processing steps were visually checked and no gross errors were found.

4.3.4. Statistics

Statistical analyses were performed using SPSS (v.13.0). Normality of the distributions was checked before parametric analyses. Percentage change over time relative to baseline was calculated as (follow-up measurement – baseline measurement) / baseline measurement x 100. Age at baseline, sex, interaction between age and sex, scanner, and time between scan acquisitions may influence relative change in brain morphology over time (Giedd et al. 1999). These variables and factors were inserted in linear regression analyses with measures of percent change in lobar brain morphology metrics as dependent variables. Out of 32 variables, a significant effect of site was found for percent change in parietal surface area only (see Appendix 4.B). No other significant main or interaction effects were found considering a p-value < 0.05 uncorrected for multiple comparisons. Therefore age at baseline, sex, interaction of age and sex, scanner, and time between acquisitions were not included in the main analyses. Thereafter three inferential analyses were conducted. First, one-sample t-tests were used to assess whether percentage change relative to baseline in CT, SA, GI, HS, WT, SD, SL and SW was significant over time (different from zero). Second, to test whether percentage change relative to baseline was invariant over the cortex we used Analyses of Variance (ANOVA) with percentage change relative to baseline for each measure as the dependent variable and lobe as the factor. Where necessary, post-hoc tests were used to determine which lobes showed significant differences. Third, Pearson partial correlation was performed to explore the direct associations among percentage change relative to baseline of CT, SA, HS, WT, SD, SL, and SW. These correlations calculate the correlation between each pair of variables while accounting for the effects of all remaining variables. In all analyses, p < 0.05 was considered significant after controlling for multiple comparisons using the False Discovery Rate (FDR) with q = 0.05 (Benjamini and Hochberg 1995). Effect size is given as Cohen‟s d.

64 The human cortex flattens during adolescence

4.3. Results

4.3.1. Longitudinal changes in cortical thickness (CT), surface area (SA), gyrification index (GI), hull surface area (HS), white matter thickness (WT), sulcal depth (SD), sulcal length (SL), and sulcal width (SW)

Over the whole cortex, CT, SA, GI, HS, SD, SL decreased over time by 1.6% (t = -5.10, p < 0.01), 1.5% (t = -7.57, p < 0.01), 1.3% (t = -8.38, p < 0.01), 0.17% (t = -0.98, p = 0.33), 0.3% (t = -1.09, p = 0.32), and 0.8% (t = -2.12, p = 0.06) relative to baseline respectively. WT and SW increased over time by 1.6% (t = 5.79, p < 0.01) and 2.5% (t = 3.80, p < 0.01) relative to baseline respectively. Table 4.3 shows the percentage change relative to baseline for all of these measures within cortical lobes and whether this change was statistically significantly after false discovery rate (FDR) correction.

Brain Lobes Measure Frontal Temporal Parietal Occipital %A tB pC % t p % t p % t p CT -1.7 -5 <.001 -1.7 -4.1 <.001 -1.4 4.3 <.001 -1.1 -2.7 0.014 SA -1.6 -6.9 <.001 -1.3 -4.1 <.001 -1.7 7.1 <.001 -0.7 -2.2 0.043 GI -1.8 -6.5 <.001 -1.2 -4.1 <.001 -0.7 1.6 0.155 -1 -3.6 0.002 HS 0.1 0.5 0.660 -0.1 -0.5 0.660 1 2.3 0.035 0.3 1.1 0.335 WT 1.7 5.5 <.001 1.2 3.6 0.002 2 5.7 <.001 1.5 3.7 <0.001 SD -0.9 -3.5 <.002 -0.6 -1.4 0.225 -0.2 0.4 0.712 -1.2 -2.9 0.008 SL -1.8 -3 0.008 0.69 -1 0.403 0.2 0.4 0.719 -1.5 -0.9 0.410 SW 3.4 4.7 <.001 -3.2 -3.1 0.006 2.3 3.1 0.008 2.2 2.7 0.013

A. Percent change relative to baseline defined as (follow-up - baseline)/baseline x 100. B. |t| value from One-Sample t test with Degrees of Freedom = 51. C. FDR-corrected p-values (q = 0.05).

Table 4.3: Percentage change relative to baseline per lobe over a two-year period in 52 healthy adolescents (age range 11-17 years).

4.3.2. Differences in longitudinal changes between lobes

ANOVAs assessed whether, for each measure, percentage change relative to baseline was different among the lobes. After FDR correction for multiple comparisons, a nearly-significant trend among lobes for a relative loss in SA (F (3,204) = 3.4, p = 0.057) and HS (F (3,204) = 3.4, p = 0.057) was found. Post-hoc tests corrected for multiple comparisons showed that the loss of SA was higher in the frontal (t (102) =

65 Chapter 4

2.6, p = 0.03, d = 0.50) and temporal cortex (t (102) = 2.8, p = 0.02, d = 0.56) compared to occipital cortex. Loss of HS was higher in the frontal cortex (t (102) = 2.2, p = 0.04, d = 0.43) and occipital cortex (t (102) = 2.5, p = 0.02, d = 0.50) compared to the temporal cortex. However, these results should be interpreted with caution because the effect of lobe was only trend-significant.

4.3.3. Partial correlations between longitudinal changes with each lobe

The patterns of association among longitudinal changes in lobar SD, SW, SA, WT and CT are illustrated by the significant partial correlations for each lobe after FDR correction for multiple comparisons (Figure 4.4). Notably, CT was positively related to SA in the frontal lobe and negatively to SW and WT in the occipital, temporal and parietal lobes.

66 The human cortex flattens during adolescence

4.4 Discussion

The present study investigated longitudinal changes in lobar cortical thickness, surface area, gyrification index, hull surface area, gyral white matter thickness, sulcal depth, sulcal length, and sulcal width using an approximately two-year measurement interval in a sample of healthy adolescents.

67 Chapter 4

The results both replicate prior findings on global cortical thinning and loss of sulcal surface area in males and females during adolescence (Raznahan et al. 2011), and extend previous findings in three important ways. First, by fractionating changes in global sulcal surface area over time into changes in lobar sulcal depth, sulcal length and sulcal width, the present study demonstrated that in addition to global sulcal widening, a decrease in sulcal depth in the frontal and occipital cortex took place over time. Second, gyral WM thickness increased over time in all lobes suggesting that previously- described lobar increases in white matter volume also take place in the gyri adjacent to the cortex (Lenroot et al. 2007; Tamnes et al. 2010). Third, the relationships among these events were compared directly, providing new insights into the dynamics of macrostructural change during adolescence as described below. Prior reports in adult and elderly participants from cross-sectional studies have described a relationship between global sulcal depth decrease and sulcal width enlargement (Magnotta et al. 1999; Im et al. 2008; Kochunov et al. 2008; Kochunov et al. 2009). Extending these studies for the first time to longitudinal adolescent brain development has revealed a direct relationship between decreased sulcal depth and sulcal widening over time that was strongest in the parietal lobe. The global widening of the sulci and decrease of depth over time constitutes a macrostructural flattening of the cortex during adolescence, most substantially in the frontal and occipital cortex (i.e., the regions where both morphological changes were significant). The simultaneous decrease in sulcal depth and increase in sulcal width over time is a neurodevelopmental process that can explain the loss of sulcal surface area, because as a sulcus becomes flatter, its surface area will also decrease. Indeed, a direct relationship between sulcal widening and loss of surface area was found in the frontal cortex. The current study also shows a global, strong, and direct relationship between sulcal widening and cortical thinning as has been shown cross-sectionally (Im et al. 2008). The decrease in lobar cortical thickness over time seems to produce a double increase in lobar sulcal width (see Table 4.3), presumably due to the presence of cortex on both sides of each sulcus (Kochunov et al. 2008). These findings point to a mechanism in which cortical thinning is related to loss of surface area via sulcal widening. The direct negative relationship between cortical thickness and surface area indicates that, after partialling out the effect of sulcal widening, individuals with the thinnest cortex have the largest surface area. This has been reported before (Seldon

68 The human cortex flattens during adolescence

2005; Hogstrom et al. 2012). It has been proposed that a negative relationship between cortical thickness and surface area is due to maturation of the white matter, expanding and stretching the outer surface like a balloon. Consequently, the outer cortical surface has to thin out to cover the expanding surface area (Seldon 2005; Hogstrom et al. 2012). An important question in adolescent brain development is whether cortical thinning reflects true atrophy or is an artifact due to the maturation of adjacent subcortical white matter (Paus 2005). Cortical thinning during adolescence may be due to underlying synaptic pruning, i.e., the use-dependent selective elimination of synapses (Rakic et al. 1994; Huttenlocher and Dabholkar 1997) together with trophic glial and vascular changes and/or cell shrinkage (Morrison et al. 1997). The concurrent presence of increasing gyral WM thickness and cortical thinning over time in the current study is coherent with prior studies showing global white matter volume increase and decrease of cortical thickness in the same age range (Lenroot et al. 2007; Shaw et al. 2008; Tamnes et al. 2010; Raznahan et al. 2011). The direct relationship between gyral WM thickness enlargement and cortical thinning in the parietal, temporal, and occipital cortex of individuals seen here provides compelling evidence that during adolescence, T1-based measurements of cortical thickness in these regions are also dependent on the maturation of the white matter in the adjacent gyri. During adolescence, white matter volume increase probably reflects underlying changes in myelination and axon diameter (Yakovlev and Lecours 1967; Benes et al. 1994; Paus 2010). Since the T1-signal is highly sensitive to changes in myelination (Walters et al. 2003), continuing myelination in the neuropil and deep intracortical layers causes voxels on the white matter / gray matter interface to be classified as white matter at later ages. Their being previously categorized as gray matter at earlier ages would lead to an apparent age-related cortical thinning (Sowell et al. 2004; Paus 2005; Shaw et al. 2008; Westlye et al. 2010; Geyer et al. 2011). The direct dependence between cortical thickness and gyral WM thickness may affect the interpretation of cortical thickness findings in subjects with ongoing white matter maturation and underscores the need to evaluate cortical thickness in the context of both gray- and white-matter development (Geyer et al. 2011; Glasser and Van Essen 2011). There are several limitations to this study which should be taken into account when interpreting the results. First, they are specific to the age range covered here; the relationships among these morphometric variables are likely to differ at other age ranges. Second, we used global and lobar measures across hemispheres which precluded

69 Chapter 4 the analysis of more regional and lateralized effects. Third, the distributions of subject numbers and males-female ratios were not homogeneous over the age range, with fewer subjects and fewer females at the onset of adolescence. Therefore, these results might be more representative of middle and late adolescence (14-18 years of age) as compared to early adolescence. Fourth, we used a multisite design but the sample size was relatively small. Fifth, developmental trajectories of cortical morphological measures may be nonlinear, depending on the type of measure and the region. The current study had only two timepoints, precluding detection of non-linear changes in morphometric variables. Future studies using these same morphometric analyses and collecting multiple measurements (> 2) over time in a sample with a wider age range and are needed to judge how accurately the present results reflect the trajectory of cortical brain changes from early to late adolescence. In conclusion, a widespread cortical thinning during adolescence to be related to sulcal widening was found. In addition, sulcal depth decreased in frontal and occipital lobes. The combination of increasing sulcal width and decreasing sulcal depth implies a flattening of the cortex which could be a mechanism producing the typical loss of cortical surface area seen during adolescence.

Acknowledgements

The authors declare no competing financial interests. The authors would like to thank all study participants and their families. This study is supported by the Instituto de Salud Carlos III, the CDTI under the CENIT Programme (AMIT Project), the Ramon y Cajal Program of the Spanish Ministry of „Economía y Competividad‟, the Fundación Alicia Koplowitz, and Caja Navarra.

70 The human cortex flattens during adolescence

Related appendices

Appendix 4.A. Manual labeling of sulci generated by the BrainVISA software into lobes

The sulcal names are from the nomenclature used by the BrainVISA software program. The BrainVISA nomenclature can be found at http://www.lnao.fr/IMG/png/BrainVISA _sulci_atlas_with_table_150dpi-r90.png

Frontal: S.F.median, S.F.median.1, S.F.median.2, S.F.median.3, S.F.median.4, S.F.sup, S.F.sup.ant, S.F.sup.ant.r.asc, S.F.sup.ant.r.desc, S.F.sup.moy, S.F.sup.moy.r.asc, S.F.sup.moy.r.desc, S.F.sup.post, S.F.sup.post.r.asc, S.F.sup.post.r.desc, S.F.inter, S.F.inter.ant, S.F.inter.ant.r.asc, S.F.inter.ant.r.desc, S.F.inter.moy, S.F.inter.moy.r.asc, S.F.inter.moy.r.desc, S.F.inter.post, S.F.inter.post.r.asc, S.F.inter.post.r.desc, S.F.inf, S.F.inf.ant, S.F.inf.ant.r.asc, S.F.inf.ant.r.desc, S.F.inf.moy, S.F.inf.moy.r.asc, S.F.inf.moy.r.desc, S.F.inf.post, S.F.inf.post.r.asc, S.F.inf.post.r.desc, S.F.inf.r.tr, S.F.polaire.tr, S.F.polaire.tr.sup, S.F.polaire.tr.moy, S.F.polaire.tr.inf, S.F.marginal, S.F.orbitaire, S.Olf, S.Or, S.Or.l, S.p.Olf.ant, S.F.int, S.F.int.pol, S.F.int.sup_, S.F.int.AMS_, S.R.sup, S.R.inf, S.p.C, S.intraCing, S.C.LPC, S.Pe.C, S.Pe.C.median, S.Pe.C.marginal, S.Pe.C.sup, S.Pe.C.sup.r.GPeC, S.Pe.C.inter, S.Pe.C.inter.r.GPeC, S.Pe.C.inf, S.Pe.C.inf.r.GPeC Parietal: S.Pa.sup, S.Pa.t, S.Po.C, S.Po.C.sup, S.Po.C.br.LPS, S.Po.C.br.LPI, S.Po.C.br.GPOC, S.GSM, F.I.P, F.I.P.Po.C, F.I.P.Po.C.sup, F.I.P.Po.C.inf, F.I.P.r.asc, F.I.P.r.trans, F.I.P.Horiz, F.I.P.ParO, F.I.P.ParO.sup, F.I.P.ParO.inf, F.I.P.r.desc, F.I.P.r.int.1, F.I.P.r.int.2, S.s.P, S.Pa.int, S.Pa.int.ant, S.Pa.int.post, F.P.O, F.P.O.sup, F.P.O.inf, S.Cu, S.Cu.sag.inf, S.Cu.sag.sup, S.Cu.vert Temporal: S.T.pol, S.T.pol.sup, S.T.pol.inf, S.ac, S.T.s, S.T.s.ant, S.T.s.ant.r.asc, S.T.s.ant.r.desc, S.T.s.moy, S.T.s.moy.r.asc, S.T.s.moy.r.desc, S.T.s.post, S.T.s.post.r.asc, S.T.s.post.r.desc, S.T.s.ter.asc.ant, S.T.s.horiz, S.T.s.ter.asc.post, S.T.post, S.T.i, S.T.i.ant, S.T.i.ant.R.ant, S.T.i.ant.R.moy, S.T.i.post, S.T.i.post.R.post, S.T.i.post.R.desc, S.O.T.marg, S.O.T.lat, S.O.T.lat.ant, S.O.T.lat.int, S.O.T.lat.med, S.O.T.lat.post, F.Coll, F.Coll.R.tr.ant, F.Coll.R.PH, F.Coll.r.ter, , F.Coll.r.IL, S.U, S.Rh, S.h Occipital: OCCIPITAL, S.O.t, S.Oc.Te.lat, S.I.O, S.O.L, S.O.L.ant, S.O.L.post, S.Lu, S.r.Lu, S.O.i, S.O.a, I.Pr.O, S.Li, S.Li.ant, S.Li.post, F.Cal.ant.-Sc.Cal

The human cortex flattens during adolescence

Appendix 4.B. Studying and removing the scanner effect over cortical metrics

As it was mentioned in this chapter, five different MRI scanners were employed to collect the study data. Different scanners produce images with different contrast levels between brain tissues which introduce a bias in the surface estimation process and thus on the metrics computed from them. As can be seen in Figure 4.B., there are some metrics that present a cross- sectional effect of site such as mean cortical thickness or gyrification index. Therefore, the relative change over time (relative to baseline) for each metric was calculated (Figure 4.B., second column). As can be appreciated, no effect of site was found for relative change in cortical thickness and gyrification index. Figure 4.B. shows that in the data no significant effect of site exists for the change over time (for all except one metric).

73 Chapter 4

74 The human cortex flattens during adolescence

75 Chapter 4

Related references

APA (1994). “Diagnostic and statistical manual of mental disorders. Washington, DC. Benes, F. M., M. Turtle, et al. (1994). "Myelination of a key relay zone in the hippocampal formation occurs in the human brain during childhood, adolescence, and adulthood." Archives of General Psychiatry 51(6): 477-484. Benjamini, Y. and Y. Hochberg (1995). "Controlling the false discovery rate: A practical and powerful approach to multiple testing." Journal of the Royal Statistical Society: Series B 57: 289-300. Blakemore, S. J. (2012). "Imaging brain development: the adolescent brain." Neuroimage 61(2): 397-406. Brouwer, R. M., R. C. Mandl, et al. (2012). "White matter development in early puberty: a longitudinal volumetric and diffusion tensor imaging twin study." PLoS One 7(4): e32316. Brown, T. T., J. M. Kuperman, et al. (2012). "Neuroanatomical assessment of biological maturity." Current Biology 22(18): 1693-1698. Buchanan, R. W. and D. W. Heinrichs (1989). "The Neurological Evaluation Scale (NES): a structured instrument for the assessment of neurological signs in schizophrenia." Psychiatry Research 27(3): 335-350. Castro-Fornieles, J., M. Parellada, et al. (2007). "The child and adolescent first-episode psychosis study (CAFEPS): design and baseline results." Schizophrenia Research 91(1- 3): 226-237. Dale, A. M., B. Fischl, et al. (1999). "Cortical surface-based analysis. I. Segmentation and surface reconstruction." Neuroimage 9(2): 179-194. Desikan, R. S., F. Segonne, et al. (2006). An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest." Neuroimage 31(3): 968-980. Fischl, B., M. I. Sereno, et al. (1999). Cortical surface-based analysis. II: Inflation, flattening, and a surface-based coordinate system." Neuroimage 9(2): 195-207. Geyer, S., M. Weiss, et al. (2011). Microstructural Parcellation of the Human Cerebral Cortex - From Brodmann's Post-Mortem Map to in vivo Mapping with High-Field Magnetic Resonance Imaging " Frontiers in Human Neuroscience 5: 19. Giedd, J. N., J. Blumenthal, et al. (1999). Brain development during childhood and adolescence: a longitudinal MRI study." Nature Neuroscience 2(10): 861-863. Glasser, M. F. and D. C. Van Essen (2011). Mapping human cortical areas in vivo based on myelin content as revealed by T1- and T2-weighted MRI." Journal of Neuroscience 31(32): 11597-11616. Gogtay, N., J. N. Giedd, et al. (2004). Dynamic mapping of human cortical development during childhood through early adulthood." Proceedings of the National Academy of Sciences 101(21): 8174-8179. Han, X., J. Jovicich, et al. (2006). "Reliability of MRI-derived measurements of human cerebral cortical thickness: the effects of field strength, scanner upgrade and manufacturer." Neuroimage 32(1): 180-194. Hogstrom, L. J., L. T. Westlye, et al. (2012). "The Structure of the Cerebral Cortex Across Adult Life: Age-Related Patterns of Surface Area, Thickness, and Gyrification." Cerebral Cortex. 23:2521-2530.

76 The human cortex flattens during adolescence

Huttenlocher, P. R. and A. S. Dabholkar (1997). "Regional differences in synaptogenesis in human cerebral cortex." Journal of Comparative Neurology 387(2): 167-178. Im, K., J. M. Lee, et al. (2008). "Sulcal morphology changes and their relationship with cortical thickness and gyral white matter volume in mild cognitive impairment and Alzheimer's disease." Neuroimage 43(1): 103-113. Jouvent, E., S. Reyes, et al. (2011). "Apathy is related to cortex morphology in CADASIL. A sulcal-based morphometry study." Neurology 76(17): 1472-1477. Kaufman, J., B. Birmaher, et al. (1997). "Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version (K-SADS-PL): initial reliability and validity data." Journal of the American Academy of Child & Adolescent Psychiatry 36(7): 980-988. Kochunov, P., D. A. Robin, et al. (2009). "Can structural MRI indices of cerebral integrity track cognitive trends in executive control function during normal maturation and adulthood?" Human Brain Mapping 30(8): 2581-2594. Kochunov, P., W. Rogers, et al. (2012). "A library of cortical morphology analysis tools to study development, aging and genetics of cerebral cortex." Neuroinformatics 10(1): 81- 96. Kochunov, P., P. M. Thompson, et al. (2008). "Relationship among neuroimaging indices of cerebral health during normal aging." Human Brain Mapping 29(1): 36-45. Kuperberg, G. R., M. R. Broome, et al. (2003). "Regionally localized thinning of the cerebral cortex in schizophrenia." Archives of General Psychiatry 60(9): 878-888. Lenroot, R. K., N. Gogtay, et al. (2007). "Sexual dimorphism of brain developmental trajectories during childhood and adolescence." Neuroimage 36(4): 1065-1073. Liu, T., W. Wen, et al. (2011). "The relationship between cortical sulcal variability and cognitive performance in the elderly." Neuroimage 56(3): 865-873. Magnotta, V. A., N. C. Andreasen, et al. (1999). "Quantitative in vivo measurement of gyrification in the human brain: changes associated with aging." Cerebral Cortex 9(2): 151-160. Mangin, J. F., D. Riviere, et al. (2004). "A framework to study the cortical folding patterns." Neuroimage 23 Suppl 1: S129-138. Morrison, S. J., N. M. Shah, et al. (1997). "Regulatory mechanisms in stem cell biology." Cell 88(3): 287-298. Paus, T. (2005). "Mapping brain maturation and cognitive development during adolescence." Trends in Cognitive Sciences 9(2): 60-68. Paus, T. (2010). "Growth of white matter in the adolescent brain: myelin or axon?" Brain and Cognition72(1): 26-35. Perrot, M., D. Riviere, et al. (2011). "Cortical sulci recognition and spatial normalization." Medical Image Analysis 15(4): 529-550. Rakic, P., J. P. Bourgeois, et al. (1994). "Synaptic development of the cerebral cortex: implications for learning, memory, and mental illness." Progress in Brain Research 102: 227-243. Raznahan, A., P. Shaw, et al. (2011). "How does your cortex grow?" Journal of Neuroscience 31(19): 7174-7177. Reuter, M. and B. Fischl (2011). "Avoiding asymmetry-induced bias in longitudinal image processing." Neuroimage 57(1): 19-21.

77 Chapter 4

Reuter, M., N. J. Schmansky, et al. (2012). "Within-subject template estimation for unbiased longitudinal image analysis." Neuroimage 61(4): 1402-1418. Ringe, W. K., K. C. Saine, et al. (2002). "Dyadic short forms of the Wechsler Adult Intelligence Scale-III." Assessment 9(3): 254-260. Rosas, H. D., A. K. Liu, et al. (2002). "Regional and progressive thinning of the cortical ribbon in Huntington's disease." Neurology 58(5): 695-701. Satler, J. (2001). "Assessment of children cognitive applications." San Diego State University Publisher Inc. Seldon, H. L. (2005). "Does brain white matter growth expand the cortex like a balloon?" Hypothesis and consequences. Laterality 10(1): 81-95. Shaw, P., N. J. Kabani, et al. (2008). "Neurodevelopmental trajectories of the human cerebral cortex." Journal of Neuroscience 28(14): 3586-3594. Sowell, E. R., P. M. Thompson, et al. (2004). "Longitudinal mapping of cortical thickness and brain growth in normal children." Journal of Neuroscience 24(38): 8223-8231. Su, S., T. White, et al. (2013). "Geometric computation of human gyrification indexes from magnetic resonance images." Hum Brain Mapp 34(5): 1230-1244. Tamnes, C. K., Y. Ostby, et al. (2010). "Brain maturation in adolescence and young adulthood: regional age-related changes in cortical thickness and white matter volume and microstructure." Cerebral Cortex 20(3): 534-548. Tamnes, C. K., K. B. Walhovd, et al. (2013). "Brain development and aging: overlapping and unique patterns of change." Neuroimage 68: 63-74. Van Essen, D. C. and H. A. Drury (1997). "Structural and functional analyses of human cerebral cortex using a surface-based atlas." Journal of Neuroscience 17(18): 7079- 7102. Van Soelen, I. L., R. M. Brouwer, et al. (2012). "Genetic influences on thinning of the cerebral cortex during development." Neuroimage 59(4): 3871-3880. Walters, N. B., G. F. Egan, et al. (2003). "In vivo identification of human cortical areas using high-resolution MRI: an approach to cerebral structure-function correlation". Proceedings of the National Academy of Sciences100(5): 2981-2986. Westlye, L. T., K. B. Walhovd, et al. (2010). "Differentiating maturational and aging-related changes of the cerebral cortex by use of thickness and signal intensity". Neuroimage 52(1): 172-185. Yakovlev, P. I. and A. R. Lecours (1967). "The myelogenetic cycles of regional maturation of the brain. Regional development of the brain in early life". A. Minkowski. Oxford, Blackwell Scientific: 3-70. Zilles, K., E. Armstrong, et al. (1988). "The human pattern of gyrification in the cerebral cortex." Anatomy and Embryology (Berl) 179(2): 173-179.

78 Cortical morphology of adolescents with bipolar disorder and with schizophrenia

Chapter 5 Cortical morphology of adolescents with bipolar disorder and with schizophrenia

Joost Janssen1, Yasser Alemán-Gómez1, Hugo Schnack, Evan Balaban, Laura Pina- Camacho, Fidel Alfaro-Almagro, Josefina Castro-Fornieles, Soraya Otero, Inmaculada Baeza, Dolores Moreno, Nuria Bargalló, Mara Parellada, Celso Arango, Manuel Desco

Schizophrenia Research. 2014 Jul;158(1-3):91-0. doi: 10.1016/j.schres.2014.06.040 1These authors contributed equally to this study.

Cortical morphology of adolescents with bipolar disorder and with schizophrenia

Abstract

Recent evidence points to overlapping decreases in cortical thickness and gyrification in the frontal lobe of patients with adult-onset schizophrenia and bipolar disorder with psychotic symptoms, but it is not clear if these findings generalize to patients with a disease onset during adolescence and what may be the mechanisms underlying a decrease in gyrification. This study analyzed cortical morphology using surface-based morphometry in 92 subjects (age range 11-18 years, 52 healthy controls and 40 adolescents with early-onset first-episode psychosis diagnosed with schizophrenia (n = 20) or bipolar disorder with psychotic symptoms (n = 20) based on a two year clinical follow up). Average lobar cortical thickness, surface area, gyrification index (GI) and sulcal width were compared between groups, and the relationship between the GI and sulcal width was assessed in the patient group. Both patients groups showed decreased cortical thickness and increased sulcal width in the frontal cortex when compared to healthy controls. The schizophrenia subgroup also had increased sulcal width in all other lobes. In the frontal cortex of the combined patient group sulcal width was negatively correlated (r = -0.58, p<0.001) with the GI. In adolescents with schizophrenia and bipolar disorder with psychotic symptoms there is cortical thinning, decreased GI and increased sulcal width of the frontal cortex present at the time of the first psychotic episode. Decreased frontal GI is associated with the widening of the frontal sulci which may reduce sulcal surface area. These results suggest that abnormal growth (or more pronounced shrinkage during adolescence) of the frontal cortex represents a shared endophenotype for psychosis.

81 Chapter 5

5.1. Introduction

Patients with early-onset first-episode psychosis (EOP), defined as the presence of psychotic symptoms within a psychiatric disorder appearing before the age of 18 years, may eventually develop schizophrenia, bipolar disorder or other psychotic disorders. By assessing cortical structural brain abnormalities in adolescent EOP patients (classified into schizophrenia or bipolar disorder with psychotic symptoms based on a two-year clinical follow-up diagnosis), we set out to elucidate whether both groups of subjects share cortical structural abnormalities at an early point in the development of the disease. Surface-based morphometry studies are able to separately measure cortical thickness, surface area and gyrification-related metrics such as the width of the sulci (Aleman-Gomez et al. 2013). Recent surface-based morphometry studies showed that, when compared to healthy controls, adult patients with schizophrenia and those with bipolar I disorder have decreased cortical thickness in lateral and medial frontal and superior temporal regions (Rimol et al. 2010). Reduced thickness and surface area extended to parietal and occipital regions in patients with schizophrenia only (Rimol et al. 2010). Rimol et al. reported no regions with cortical thinning that were uniquely affected in bipolar I disorder and overlapping regions in the frontal cortex such as the medial and middle superior frontal gyrus showed greater thinning in schizophrenia when compared to bipolar I disorder (Rimol et al. 2010). Decreased gyrification has also been demonstrated in the frontal cortex in both disorders and, as with cortical thickness, seems more pronounced in schizophrenia (Palaniyappan and Liddle 2014). A reduction in gyrification, i.e., a reduction in cortical surface complexity, may be partly due to increased sulcal width (Kochunov et al. 2005; Hogstrom et al. 2012; Liu et al. 2012; Zilles et al. 2013). However, a direct relationship between sulcal width and gyrification has not yet been assessed in psychotic populations. Taken together, it is unclear if the cortical surface-based findings from studies in adults can be extrapolated to adolescent-onset schizophrenia and bipolar disorder, as no previous surface-based morphometry study has compared adolescent-onset schizophrenia, bipolar disorder with psychotic symptoms and healthy controls. This study simultaneously examines lobar cortical thickness, surface area, the gyrification index (GI) and sulcal width in healthy controls, EOP-schizophrenia and EOP-bipolar disorder with psychotic symptoms subgroups. Based on studies in adults

82 Cortical morphology of adolescents with bipolar disorder and with schizophrenia we hypothesized, first, that both patient subgroups would show decreased cortical thickness and GI and increased sulcal width in the frontal lobe with respect to controls. Second, we hypothesized that an increased sulcal width would be associated with a decreased GI in patients with EOP.

5.2 Materials and methods

5.2.1. Subjects sample

The sample was selected out of a larger group of EOP patients and controls belonging to two studies (Moreno et al. 2005; Castro-Fornieles et al. 2007) of which one is a multicenter study (Castro-Fornieles et al. 2007). Inclusion criteria for patients were: 1) age between 7 and 17 years, 2) presenting with a first psychotic episode following a DSM-IV disorder of less than six months duration, and 3) absence of concomitant Axis I disorder at the time of evaluation (APA 1994). Control subjects were recruited from the same schools and residential areas as the patients. The inclusion criteria for healthy controls were: 1) age between 7 and 17 years and 2) having no psychiatric diagnosis. The exclusion criteria for both groups were: 1) mental retardation according to DSM-IV-TR criteria (APA 1994) (not only an intelligence quotient (IQ) below 70 but also impaired functioning), 2) any neurological or pervasive developmental disorder, 3) history of head trauma with loss of consciousness, 4) pregnancy and 5) substance abuse or dependence but not use if psychotic symptoms persisted 14 days after a negative urine drug test result. Only those patients and healthy controls that had a high quality MRI assessment at baseline and a valid clinical and diagnostic assessment at baseline and at the two-year follow-up visit were eligible for inclusion in the study. For the present analyses we only included patients diagnosed with schizophrenia or bipolar disorder at the two-year follow-up visit. To account for the lack of specificity of clinical presentation of first episodes of psychosis at early stages, and in order to avoid the problem of diagnostic instability, we used the 2-year follow-up diagnoses. Due to the nature of the inclusion criteria all the bipolar patients are type I with psychotic symptoms, which may lessen the psychopathological differences as measured by clinical symptom severity scales between the schizophrenia and bipolar group (see Table 1). In addition, presence of not mood congruent psychopathology and depressive symptoms is quite frequent in early-onset schizophrenia as well as early-onset bipolar disorder (Rapado-Castro et al. 2010). We

83 Chapter 5 have previously shown the diagnostic stability of both schizophrenia and bipolar disorder after the first 2-years of follow-up (Fraguas et al. 2008; Castro-Fornieles et al. 2011). A total of 94 patients and 78 controls fulfilled these criteria (see Figure 5.A.1 in Appendix 5.A). Due to the fine-grained nature of the image-analyses, only subjects with high quality baseline T1 MRI images were included in the study (see Image quality control). Excluded patients were significantly younger compared to included patients and there were no differences in other demographic, functional and clinical characteristics (Table 5.A.1 in Appendix 5.A). The final study sample included 40 adolescents with a first-episode of psychosis (20 with schizophrenia, 20 with bipolar disorder with psychotic symptoms) and 52 healthy controls, aged between 11-18 years. Patients and controls were recruited from five sites, the ratios of patients: controls were not significantly different between sites, site 1: 25:26, site 2: 4:4, site 3: 3:2, site 4: 1:6, site 5: 7:14 (Fisher‟s exact test, p = 0.34). The study was approved by the institutional review boards of the participating clinical centers. After the study was thoroughly explained to the participants, written informed consent was obtained from both the legal representatives and the individuals (if older than 12 years of age). All participants met MRI safety criteria.

5.2.2. Clinical, functional and cognitive assessment

For patients and controls, recording of sociodemographic data, physical examination, diagnostic, clinical, and functional assessments were performed by trained psychiatrists and cognitive assessments were performed by trained psychologists at different times at the corresponding clinical center. The psychiatrist was the same for each patient and control at baseline and at the 2-year clinical follow-up assessment. Diagnosis was established according to DSM-IV criteria (APA 1994), using the Spanish version of the Kiddie-Schedule for Affective Disorders and Schizophrenia, Present and Lifetime Version, a semi-structured diagnostic interview designed to assess current and past psychopathologic conditions (Kaufman et al. 1997). Diagnostic consensus was achieved during a consensus meeting between two psychiatrists for those patients for whom the diagnosis was in doubt. A stable K-SADS diagnosis at the 2-year clinical follow-up assessment was used for patient classification. Severity of symptoms at baseline was measured using the Spanish validated version of the Positive and Negative Syndrome Scale (PANSS) (Kay et al. 1987). The reliability of the different psychiatrists administering the scale was evaluated and within-class

84 Cortical morphology of adolescents with bipolar disorder and with schizophrenia correlations were higher than 0.8. Severity of manic and depressive episodes at baseline were measured using the Young Mania Rating Scale (Young et al. 1978) and Hamilton Depression Rating scale (Hamilton 1960) respectively. Severity of disability and level of functioning were determined using the Children‟s Global Assessment of Functioning (C-GAF) scale (Shaffer et al. 1983). Intellectual functioning was estimated using the Vocabulary, Information, and Block Design subtests from the Spanish version of the Wechsler Intelligence Scale for Children - Revised (WISC - R). IQ was estimated from these 3 subtests following (Silverstein 1985; Ringe et al. 2002) and is reported to show good correspondence with full-scale IQ (Satler 2001). Inter-rater reliability determined using the intraclass correlation coefficient ranged from 0.80 to 0.99 on all administered scales. Cognitive assessments were performed between weeks 4 and 8 after recruitment, either at the end of inpatient care or early in the course of outpatient treatment. Handedness was assessed by means of item 5 of the Neurological Evaluation Scale (Buchanan and Heinrichs 1989). The date of onset of first positive symptoms was assessed by means of a psychiatric interview with the patient and relatives and a review of medical records. The age at onset of illness was defined as the time between the date of birth and the date of onset of first positive symptoms (i.e., delusions or hallucinations), the mean duration of illness was defined as the time between onset of the first positive symptom and scan acquisition. Chlorpromazine equivalents were used to derive the cumulative antipsychotic medication intake at scan acquisition (Rijcken et al. 2003; Andreasen et al. 2010).

5.2.3. MRI acquisition

An anatomical brain MRI scan was acquired using five different 1.5T MRI scanners (2 Siemens Symphony scanners; 2 General Electric Signa scanners; and 1 Philips ACS Gyroscan). Data were collected at each centre and processed at one site only. The acquisition protocol consisted of a T1-weighted, three-dimensional, gradient echo sequence with 1.5 mm slice thickness (in-plane voxel size 0.98 x 0.98 mm2), and a T2-weighted turbo spin echo sequence with 3.5 mm slice thickness (in-plane voxel size 0.98 x 0.98 mm2). Full details about the acquisition parameters and interrater reliability between scanners can be found elsewhere (Reig et al. 2009).

85 Chapter 5

Both T1- and T2-weighted images were used for clinical neurodiagnostic evaluation by an independent neuroradiologist. No participants showed clinically significant brain pathology.

5.2.4. Image analysis

Image quality control As in previous studies (Aleman-Gomez et al. 2013; Nenadic et al. 2014) (see Chapter 4), image quality was determined with the support of two tools. 1) The “Check sample homogeneity” tool in the SPM-VBM8 toolbox (v.r435, http://dbm.neuro.uni- jena.de/vbm/check-sample-homogeneity/). This tool calculates the standard deviation by the sum of the squared distance of each image from the sample mean. Images that were more than 1.5 standard deviations from the mean were checked visually and excluded if deemed of insufficient quality. 2) The FreeSurfer QA tool (v5.1, http://surfer.nmr.mgh.harvard.edu/fswiki/QATools), this tool generates snapshots of anatomically labeled surfaces which were checked for topological defects and high label accuracy. Images were excluded if deemed of insufficient quality.

Intracranial volume assessment Intracranial volume (ICV) was estimated from skull-stripped baseline images that were processed using SPM8 as explained in the previous chapter (Aleman-Gomez et al. 2013).

Lobar cortical thickness, surface area, the GI, sulcal width The default FreeSurfer stream (v5.1, http://surfer.nmr.mgh.harvard.edu/) was used to generate lobar cortical thickness and lobar pial surface area. The lobar GI and sulcal width were assessed as explained in chapter 4 (Aleman-Gomez et al. 2013). All of the measurements were done in the native space of the subject‟s images. For each subject, all of the image processing steps were visually checked and no gross errors were found.

5.2.5. Statistics

Statistical analyses were performed using SPSS (v.13.0). To test for group differences in demographic and clinical variables, Analysis of Variance with post hoc tests for continuous variables were used. Chi square tests were used for discrete categorical variables. To reduce the number of tests, the measures of cortical morphology were averaged over the two hemispheres. Age, sex, site, handedness and

86 Cortical morphology of adolescents with bipolar disorder and with schizophrenia

ICV may influence brain morphology. For cortical thickness and the GI, the relationship with ICV is not clear (Rakic 1988). The results for cortical thickness and the GI did not change after including ICV as a covariate; the results with controlling for ICV are reported here. Cortical metrics were thus adjusted for age, sex, site, handedness and ICV using multiple regression analysis. In order to assess the effect of site we plotted all unstandardized residuals by lobe (see Figure 5.1). As can be seen in Figure 5.1, subjects were color coded for site. The conglomeration of colors for all measures, i.e., the lack of a clear grouping of colors, suggests that effectively after controlling for site no effect of site was present in the data. In addition, to further explore the effect of site for each diagnostic subgroup we ran a linear regression analysis with site as the predictor variable and the unadjusted (raw) value of each lobar measure of cortical morphology as the dependent variable. The p-values of these analyses are also shown in Figure 5.1. As can be seen in Figure 5.1 no significant effect of site was found (p < 0.05) for all subgroups except for cortical thickness in the occipital lobe of the schizophrenia subgroup (see Discussion). Thus, residuals (adjusted for age, sex, site, handedness and ICV) were used in the group comparisons and correlation analyses to assess the a priori hypotheses. Estimated IQ was not included as a regressor since decreased cognitive performance is itself a feature of schizophrenia and bipolar disorder. In previous studies in children and adolescents with a first psychotic episode, a cognitive impairment that does not distinguish between those with schizophrenia and bipolar disorder, with both groups scoring worse than healthy controls in all assessed cognitive domains is observed (Arango et al. 2012; Bombin et al. 2013). In addition, recent meta-analyses demonstrated strong evidence for low IQ as a putative antecedent for schizophrenia (Woodberry et al. 2008; Matheson et al. 2011). Other studies have also reported low IQ in patients with psychosis (Goldberg et al. 1988; Kenny et al. 1997; Tiihonen et al. 2005; Lewandowski et al. 2011). The General Linear Model is based on a number of assumptions, including that the observed values have a linear, additive structure, that the residuals of the model fit have the same variance and are normally distributed. Not all variables in the presented study met all these requirements. In addition, measurements related to biological morphology, such as lengths, areas, volumes or weights, are well known to follow non-normal distributions (Winkler et al. 2012). To avoid the potential influence of non-normality and heteroscedasticity on the analyses of morphological data all tests involving morphological data were done non-parametrically. To test whether group differences in lobar cortical thickness, surface area, GI and sulcal width existed,

87 Chapter 5 pair-wise comparisons between controls and schizophrenia, controls and bipolar disorder with psychotic symptoms, and schizophrenia and bipolar disorder with psychotic symptoms were conducted using Mann–Whitney U tests. To control for multiple comparisons we used the False Discovery Rate (FDR) with q = 0.05 (Benjamini and Hochberg 1995) on all the output p-values from the group comparisons (n=48). To assess whether increased sulcal width was associated with lower frontal GI in early-onset psychosis, Spearman rank correlations were used to test the association between frontal GI and sulcal width. Morphological variables were also correlated with clinical variables and accumulative antipsychotic medication use in chlorpromazine equivalents.

88 Cortical morphology of adolescents with bipolar disorder and with schizophrenia

5.3. Results

5.3.1. Demographic and clinical differences

There were no significant group differences in age, sex, handedness, and parental socioeconomic status (see Table 5.1). As per Table 5.1, parental education and estimated IQ were lower in patients with schizophrenia and bipolar disorder compared to controls, and there were no significant differences between patients with schizophrenia and bipolar disorder.

89 Chapter 5

90 Cortical morphology of adolescents with bipolar disorder and with schizophrenia

year year

symptoms.

onset onset psychosis (EOP) patients with a two

up diagnosis of bipolar disorder with psychotic with psychotic disorder bipolar of diagnosis up

year follow year

Sociodemographic Sociodemographic and clinical characteristics at baseline of healthy controls, early

up diagnosis of schizophrenia and EOP patients with a 2 a with patients EOP and schizophrenia of diagnosis up

Table Table 5. follow

91 Chapter 5

5.3.2. Group differences in cortical thickness, surface area, gyrification index, and sulcal width

Table 5.2 shows the unadjusted mean values and the adjusted median values for the lobar measurements and whether the adjusted median values were significantly different between groups before and after correction for multiple comparisons (significant p- values after correction for multiple comparisons in bold). Scatterplots for the five measurements that differed significantly between groups after correction for multiple comparisons are shown in Figure 5.2. Leaving out the patients that used lithium did not significantly alter the results.

Figure 5.2: Scatterplots showing significant differences after correction for multiple comparisons in cortical morphological measurements between healthy controls (n=52), early- onset psychosis (EOP) patients with a two-year follow-up diagnosis of schizophrenia (n=20) and EOP patients with a 2-year follow-up diagnosis of bipolar disorder with psychotic symptoms (n=20). Values are adjusted for age, sex, site, handedness and ICV, solid bar represents the median. C=controls; SCZ=schizophrenia; BD=bipolar disorder.

92 Cortical morphology of adolescents with bipolar disorder and with schizophrenia

up diagnosis up diagnosis of

year year follow

onset psychosis (EOP) onset psychosis patients (EOP) with a two

up diagnosis of bipolar disorder with psychotic symptoms (n=20) symptoms psychotic with disorder ofbipolar diagnosis up

year follow year

wise comparisons wise between controls healthy comparisons early (n=52),

Pair

Table Table 5. 2 a with patients EOP and (n=20) schizophrenia

93 Chapter 5

5.3.3. Correlation between frontal gyrification index and sulcal width in the patient group

After adjusting for age, sex, site, ICV and handedness, the Spearman‟s rho revealed a statistically significant relationship between the mean sulcal width and the GI of the frontal lobe (r=-0.58, p<0.001) (see Figure 5.3).This meant that those patients with a larger sulcal opening had a lower GI in the frontal lobe.

5.3.4. Correlations between the morphometry measurements and clinical variables

There were no significant correlations between the morphometry measurements, cumulative antipsychotic medication intake (chlorpromazine equivalents) and PANSS and c-GAF subscales.

5.4. Discussion

This work compared the cortical morphology of adolescents with first-episode early-onset psychosis diagnosed with schizophrenia or bipolar disorder with psychotic symptoms (according to a two-year follow-up diagnosis) and healthy subjects using novel surface-based morphometric techniques. The main findings of the study were, first, a reduced frontal cortical thickness and an increased sulcal width in both patient subgroups and an increased temporal, parietal, and occipital sulcal width in the schizophrenia subgroup when compared to controls. Second, both patient subgroups

94 Cortical morphology of adolescents with bipolar disorder and with schizophrenia showed a decreased frontal GI when compared to controls although these differences did not survive correction for multiple comparisons. Three, an increase in frontal sulcal width was associated with a lower frontal GI in the combined patient group. An innovative aspect of the current study is the assessment of the association between sulcal width and the GI. Irrespective of age, sex, site, brain size and handedness we found that patients with increased sulcal width had a lower GI. This could be part of a mechanism explaining the lower frontal GI in both patient groups. Prior studies investigating senescence have shown that increasing sulcal width is associated with a reduction in surface area and a lower gyrification (Magnotta et al. 1999; Kochunov et al. 2009; Hogstrom et al. 2012). Results point to a similar relationship in adolescent brain development, i.e., wider sulci may have a reduced surface area and thus contribute to a lower GI in both patient subgroups. These findings are consistent with decreased frontal cortical thickness and gyrification in childhood and early-onset schizophrenia, adult-onset first-episode psychosis and chronic schizophrenia (White et al. 2003; Narr et al. 2005; Greenstein et al. 2006; Voets et al. 2008; Rimol et al. 2010; Palaniyappan et al. 2011) the best of our knowledge, this is the first study that assessed sulcal span in childhood and adolescent psychotic populations. Another aspect of sulcal morphology, sulcal depth, was not abnormal in the anterior cingulate sulcus in adults with first-episode psychosis and schizophrenia (Csernansky et al. 2008; Fornito et al. 2008; Rametti et al. 2010). With regard to bipolar disorder, reduced frontal cortical thickness and gyrification has been demonstrated in adult patients (McIntosh et al. 2009; Penttila et al. 2009; Foland-Ross et al. 2011; Palaniyappan et al. 2011) but findings for adolescent-onset bipolar disorder have been inconclusive (Wilke et al. 2004; Chang et al. 2005; Dickstein et al. 2005; Frazier et al. 2005; Kaur et al. 2005; Sanches et al. 2005; Adler et al. 2007). The current study sample differs in two important ways from previously studied adolescent-onset bipolar populations which may have influenced the results. First, all the patients were psychotic at the time of study intake. Second, all patients were inpatients at the time of their first psychotic episode. Prior studies have used mixed samples of in- and out- patients and patients, allowing for the possibility that different degrees of illness severity and different mood states at intake modulate brain morphology (Nery et al. 2009). In this study, all of the patients were on anti-psychotic medication at the time of scan acquisition and four of patients with a follow-up diagnosis of bipolar disorder had treatment with lithium at the time of scan acquisition. The inclusion of patients treated

95 Chapter 5 with lithium can be counterproductive for brain morphology studies assessing diagnostic effects since treatment with lithium may lead to structure enlargement (Ho and Magnotta 2010; Cousins et al. 2013) canceling out a group diagnostic effect. However, treatment duration (defined as the time between study enrolment and scan acquisition) was 28 and 21 days for the schizophrenia and bipolar subgroups respectively. We deem it therefore unlikely that medication intake biased these results. Indeed, antipsychotic medication intake did not correlate significantly with any of the structural brain measures. For decreases of cortical thickness in the temporal, parietal and occipital cortices, the schizophrenia subgroup differed more strongly from controls than the bipolar subgroup did. The schizophrenia subgroup showed reduced cortical thickness in the parietal cortex when compared to the controls but this difference was no longer significant when corrected for multiple comparisons. Therefore this finding should be interpreted with caution. Nevertheless, reduction in parietal cortical thickness in adult-onset schizophrenia has been reported (Rimol et al. 2010). In addition, temporal, parietal, and occipital sulcal width of the control and bipolar subgroups was similar and therefore both differed from the schizophrenia subgroup although the difference in sulcal width between the patient subgroups did not survive correction for multiple comparisons. These results may be indicative of more diffuse aberrant cortical morphology in the schizophrenia subgroup as compared to the bipolar subgroup. Studies in adults have shown more widespread cortical thinning in schizophrenia when compared to bipolar disorder with psychotic symptoms (Rimol et al. 2010) adding support for the notion that schizophrenia may affect more diffuse areas of the cortex compared to bipolar disorder with psychotic symptoms. In addition to decreased frontal cortical thickness, fMRI studies have demonstrated abnormal frontal brain activation and cognitive impairments in working memory, response inhibition, and goal-directed behavior suggesting that abnormal frontal cortical structure and function are hallmark characteristics of schizophrenia and bipolar disorder (Harrison 1999; Owen et al. 1999; Martinez-Aran et al. 2008; Pomarol-Clotet et al. 2010; Lim et al. 2013). These findings, together with the prior literature, suggest that at an early phase in schizophrenia there may be involvement of dismorphologies in multiple distributed brain regions, while for bipolar disorder with psychotic symptoms this is confined to the frontal regions (van Haren et al. 2011; Rimol et al. 2012). What does the thinner cortex in schizophrenia and bipolar disorder represent? Cortical thinning during adolescence may be due to underlying synaptic pruning, i.e.,

96 Cortical morphology of adolescents with bipolar disorder and with schizophrenia the use-dependent selective elimination of synapses together with trophic glial and vascular changes and/or cell shrinkage (Rakic et al. 1994; Huttenlocher and Dabholkar 1997). The thinner frontal cortices in adolescent patients with early-onset psychosis suggest that abnormal growth (or more pronounced shrinkage during adolescence) of the frontal cortex represents a shared endophenotype for psychosis, consistent with prior reports of progressive frontal volume loss in adolescent-onset schizophrenia and bipolar disorder with psychotic symptoms (Arango et al. 2012) and thinning of these regions in normal childhood and adolescent brain development (Shaw et al. 2008). There are several limitations to this study which should be taken into account when interpreting the results. Firstly, we did not have access to full-scale premorbid and current IQ for the patients. Patients underwent extensive neuropsychological testing. To reduce the time of assessment, we estimated IQ on the basis of two Wechsler Adult Intelligence Scale subtests. Secondly, brain changes due to antipsychotic treatment cannot be ruled out. However, a strength of the current study is the short mean duration of the first psychotic episode. Furthermore, no correlation within patients between anatomical variation and dose of chlorpromazine equivalents used was found. Thirdly, we did not have information on pubertal status which may confound the assessment of brain morphology. Fourthly, due to the multisite study approach, the results could have been influenced by differences in MR scanner type. Figure 5.1shows that the effect of site was minimized by using the residuals for all measures and subgroups except the cortical thickness of the occipital lobe in the schizophrenia subgroup. This means that the results for the subgroup comparisons of occipital cortical thickness have to be interpreted with caution. Fifthly, we analyzed cortical morphology at the lobar level which does not exclude the possibility of differences in other areas of the cortex. In conclusion, a decreased frontal cortical thickness and increased sulcal width in adolescents with psychoses that were diagnosed with schizophrenia or bipolar disorder with psychotic symptoms after a two-year clinical follow up when compared to controls was found. Increased sulcal width was associated with a lower frontal GI in the combined patient group. In adolescents with schizophrenia, sulcal width of the temporal, parietal and occipital cortices were also increased suggesting that at an early phase in schizophrenia there is involvement of multiple distributed brain regions, while for bipolar disorder with psychotic symptoms cortical involvement is more confined to the frontal regions.

97 Chapter 5

Related appendices

Appendix 5.A. Quality control and final sample

The Figure 5.A.1 constitutes flowchart about the followed quality control process to get the final study sample. Different clinical and technical criterions were employed to define the final number of subjecs for both patients and healthy controls. Sociodemographic and clinical variables at baseline of included and excluded healthy controls and early-onset psychosis (EOP) patients are displayed at Table 5.A.1.

Figure 5.A.1: Quality control process to close the final study sample (green rectangle). Clinical or technical exclusion criterions as well as the number of excluded subjects are also displayed.

Healthy Controls Patients

Included Excluded p Included Excluded p (N=52) (N=26) (N=40) (N=54)

Age (years) 15.4 (1.4) 15.4 (1.7) .98 16.2 (1.4) 15.2 (2.0) <.01 Sex, n (% male) 32 (62%) 21 (75%) .14 31 (77%) 37 (67%) .47 Estimated IQ 108 (17.6) 106 (17.2) .63 80 (17.4) 83 (16.9) .40

Symptoms at baseline PANSS positive 26.2 (4.8) 24.4 (6.3) .12 PANSS negative 20.8 (9.2) 21.8 (8.6) .59 PANSS general 45.8 (9.8) 46.4 (11.6) .79 c-GAF 35.3 (16.0) 36.5 (17.0) .73

IQ = intelligence quotient PANSS = Positive and Negative Symptoms Scale c-GAF = Children’s Global Assessment of Functioning scale

Table 5.A.1: Sociodemographic and clinical characteristics at baseline of included and excluded healthy controls and early-onset psychosis (EOP) patients.

98 Cortical morphology of adolescents with bipolar disorder and with schizophrenia

Related references

Adler, C. M., M. P. DelBello, et al. (2007). "Voxel-based study of structural changes in first- episode patients with bipolar disorder." Biological Psychiatry 61(6): 776-781. Aleman-Gomez, Y., J. Janssen, et al. (2013). "The human cerebral cortex flattens during adolescence." Journal of Neuroscience 33(38): 15004-15010. Andreasen, N. C., M. Pressler, et al. (2010). "Antipsychotic dose equivalents and dose-years: a standardized method for comparing exposure to different drugs." Biological Psychiatry 67(3): 255-262. APA (1994). "Diagnostic and statistical manual of mental disorders." Washington, DC. Arango, C., M. Rapado-Castro, et al. (2012). "Progressive brain changes in children and adolescents with first-episode psychosis." Archives of General Psychiatry 69(1): 16-26. Benjamini, Y. and Y. Hochberg (1995). "Controlling the false discovery rate: A practical and powerful approach to multiple testing." Journal of the Royal Statistical Society. Series B 57(1): 289-300. Bombin, I., M. Mayoral, et al. (2013). "Neuropsychological evidence for abnormal neurodevelopment associated with early-onset psychoses." Psychological Medicine 43(4): 757-768. Buchanan, R. W. and D. W. Heinrichs (1989). "The Neurological Evaluation Scale (NES): a structured instrument for the assessment of neurological signs in schizophrenia." Psychiatry Research 27(3): 335-350. Castro-Fornieles, J., I. Baeza, et al. (2011). "Two-year diagnostic stability in early-onset first- episode psychosis." Journal of Child Psychology and Psychiatry 52(10): 1089-1098. Castro-Fornieles, J., M. Parellada, et al. (2007). "The child and adolescent first-episode psychosis study (CAFEPS): design and baseline results." Schizophrenia Research 91(1- 3): 226-237. Cousins, D. A., B. Aribisala, et al. (2013). "Lithium, gray matter, and magnetic resonance imaging signal." Biological Psychiatry 73(7): 652-657. Csernansky, J. G., S. K. Gillespie, et al. (2008). "Symmetric abnormalities in sulcal patterning in schizophrenia." Neuroimage 43(3): 440-446. Chang, K., N. Barnea-Goraly, et al. (2005). "Cortical magnetic resonance imaging findings in familial pediatric bipolar disorder." Biological Psychiatry 58(3): 197-203. Dickstein, D. P., M. P. Milham, et al. (2005). "Frontotemporal alterations in pediatric bipolar disorder: results of a voxel-based morphometry study." Archives of General Psychiatry 62(7): 734-741. Foland-Ross, L. C., P. M. Thompson, et al. (2011). "Investigation of cortical thickness abnormalities in lithium-free adults with bipolar I disorder using cortical pattern matching." American Journal of Psychiatry 168(5): 530-539. Fornito, A., M. Yucel, et al. (2008). "Surface-based morphometry of the anterior cingulate cortex in first episode schizophrenia." Human Brain Mapping 29(4): 478-489. Fraguas, D., M. J. de Castro, et al. (2008). "Does diagnostic classification of early-onset psychosis change over follow-up?" Child Psychiatry & Human Development 39(2): 137-145. Frazier, J. A., J. L. Breeze, et al. (2005). "Cortical gray matter differences identified by structural magnetic resonance imaging in pediatric bipolar disorder." Bipolar Disorders 7(6): 555-569.

99 Chapter 5

Goldberg, T. E., C. N. Karson, et al. (1988). "Intellectual impairment in adolescent psychosis. A controlled psychometric study." Schizophrenia Research 1(4): 261-266. Greenstein, D., J. Lerch, et al. (2006). "Childhood onset schizophrenia: cortical brain abnormalities as young adults." Journal of Child Psychology and Psychiatry 47(10): 1003-1012. Hamilton, M. (1960). "A rating scale for depression." Journal of Neurology, Neurosurgery & Psychiatry 23: 56-62. Harrison, P. J. (1999). "The neuropathology of schizophrenia. A critical review of the data and their interpretation." Brain 122 ( Pt 4): 593-624. Ho, B. C. and V. Magnotta (2010). "Hippocampal volume deficits and shape deformities in young biological relatives of schizophrenia probands." Neuroimage 49(4): 3385-3393. Hogstrom, L. J., L. T. Westlye, et al. (2012). "The Structure of the Cerebral Cortex Across Adult Life: Age-Related Patterns of Surface Area, Thickness, and Gyrification." Cerebral Cortex. 23:2521-2530. Huttenlocher, P. R. and A. S. Dabholkar (1997). "Regional differences in synaptogenesis in human cerebral cortex." Journal of Comparative Neurology 387(2): 167-178. Kaufman, J., B. Birmaher, et al. (1997). "Schedule for Affective Disorders and Schizophrenia for School-Age Children-Present and Lifetime Version (K-SADS-PL): initial reliability and validity data." Journal of the American Academy of Child & Adolescent Psychiatry 36(7): 980-988. Kaur, S., R. B. Sassi, et al. (2005). "Cingulate cortex anatomical abnormalities in children and adolescents with bipolar disorder." American Journal of Psychiatry 162(9): 1637-1643. Kay, S. R., A. Fiszbein, et al. (1987). "The positive and negative syndrome scale (PANSS) for schizophrenia." Schizophrenia Bulletin 13(2): 261-276. Kenny, J. T., L. Friedman, et al. (1997). "Cognitive impairment in adolescents with schizophrenia." American Journal of Psychiatry 154(11): 1613-1615. Kochunov, P., J. F. Mangin, et al. (2005). "Age-related morphology trends of cortical sulci." Human Brain Mapping 26(3): 210-220. Kochunov, P., D. A. Robin, et al. (2009). "Can structural MRI indices of cerebral integrity track cognitive trends in executive control function during normal maturation and adulthood?" Human Brain Mapping 30(8): 2581-2594. Lewandowski, K. E., B. M. Cohen, et al. (2011). "Relationship of neurocognitive deficits to diagnosis and symptoms across affective and non-affective psychoses." Schizophrenia Research 133(1-3): 212-217. Lim, C. S., R. J. Baldessarini, et al. (2013). "Longitudinal neuroimaging and neuropsychological changes in bipolar disorder patients: review of the evidence." Neuroscience & Biobehavioral Reviews 37(3): 418-435. Liu, T., D. M. Lipnicki, et al. (2012). "Cortical gyrification and sulcal spans in early stage Alzheimer's disease." PLoS One 7(2): e31083. Magnotta, V. A., N. C. Andreasen, et al. (1999). "Quantitative in vivo measurement of gyrification in the human brain: changes associated with aging." Cerebral Cortex 9(2): 151-160. Martinez-Aran, A., C. Torrent, et al. (2008). "Neurocognitive impairment in bipolar patients with and without history of psychosis." Journal of Clinical Psychiatry 69(2): 233-239.

100 Cortical morphology of adolescents with bipolar disorder and with schizophrenia

Matheson, S. L., A. M. Shepherd, et al. (2011). "A systematic meta-review grading the evidence for non-genetic risk factors and putative antecedents of schizophrenia." Schizophrenia Research 133(1-3): 133-142. McIntosh, A. M., T. W. Moorhead, et al. (2009). "Prefrontal gyral folding and its cognitive correlates in bipolar disorder and schizophrenia." Acta Psychiatrica Scandinavica 119(3): 192-198. Moreno, D., M. Burdalo, et al. (2005). "Structural neuroimaging in adolescents with a first psychotic episode." Journal of the American Academy of Child & Adolescent Psychiatry 44(11): 1151-1157. Narr, K. L., R. M. Bilder, et al. (2005). "Mapping cortical thickness and gray matter concentration in first episode schizophrenia." Cerebral Cortex 15(6): 708-719. Nenadic, I., R. A. Yotter, et al. (2014). "Cortical surface complexity in frontal and temporal areas varies across subgroups of schizophrenia." Human Brain Mapping 35(4): 1691- 1699. Nery, F. G., H. H. Chen, et al. (2009). "Orbitofrontal cortex gray matter volumes in bipolar disorder patients: a region-of-interest MRI study." Bipolar Disorders 11(2): 145-153. Owen, A. M., N. J. Herrod, et al. (1999). "Redefining the functional organization of working memory processes within human lateral prefrontal cortex." European Journal of Neuroscience 11(2): 567-574. Palaniyappan, L. and P. F. Liddle (2014). "Diagnostic discontinuity in psychosis: a combined study of cortical gyrification and functional connectivity." Schizophrenia Bulletin 40(3): 675-684. Palaniyappan, L., P. Mallikarjun, et al. (2011). "Folding of the prefrontal cortex in schizophrenia: regional differences in gyrification." Biological Psychiatry 69(10): 974- 979. Penttila, J., A. Cachia, et al. (2009). "Cortical folding difference between patients with early- onset and patients with intermediate-onset bipolar disorder." Bipolar Disorders 11(4): 361-370. Pomarol-Clotet, E., E. J. Canales-Rodriguez, et al. (2010). "Medial prefrontal cortex pathology in schizophrenia as revealed by convergent findings from multimodal imaging." Molecular Psychiatry 15(8): 823-830. Rakic, P. (1988). "Specification of cerebral cortical areas." Science 241(4862): 170-176. Rakic, P., J. P. Bourgeois, et al. (1994). "Synaptic development of the cerebral cortex: implications for learning, memory, and mental illness." Progress in Brain Research 102: 227-243. Rametti, G., C. Junque, et al. (2010). "Anterior cingulate and paracingulate sulci morphology in patients with schizophrenia." Schizophrenia Research 121(1-3): 66-74. Rapado-Castro, M., C. Soutullo, et al. (2010). "Predominance of symptoms over time in early- onset psychosis: a principal component factor analysis of the Positive and Negative Syndrome Scale." Journal of Clinical Psychiatry 71(3): 327-337. Reig, S., J. Sanchez-Gonzalez, et al. (2009). "Assessment of the increase in variability when combining volumetric data from different scanners." Human Brain Mapping 30(2): 355- 368. Rijcken, C. A., T. B. Monster, et al. (2003). "Chlorpromazine equivalents versus defined daily doses: how to compare antipsychotic drug doses?" Journal of Clinical Psychopharmacology 23(6): 657-659.

101 Chapter 5

Rimol, L. M., C. B. Hartberg, et al. (2010). "Cortical thickness and subcortical volumes in schizophrenia and bipolar disorder." Biological Psychiatry 68(1): 41-50. Rimol, L. M., R. Nesvag, et al. (2012). "Cortical volume, surface area, and thickness in schizophrenia and bipolar disorder." Biological Psychiatry 71(6): 552-560. Ringe, W. K., K. C. Saine, et al. (2002). "Dyadic short forms of the Wechsler Adult Intelligence Scale-III." Assessment 9(3): 254-260. Sanches, M., R. B. Sassi, et al. (2005). "Subgenual prefrontal cortex of child and adolescent bipolar patients: a morphometric magnetic resonance imaging study." Psychiatry Research 138(1): 43-49. Satler, J. (2001). "Assessment of children cognitive applications." San Diego State University Publisher Inc. Shaffer, D., M. S. Gould, et al. (1983). "A children's global assessment scale (CGAS)." Archives of General Psychiatry 40(11): 1228-1231. Shaw, P., N. J. Kabani, et al. (2008). "Neurodevelopmental trajectories of the human cerebral cortex." Journal of Neuroscience 28(14): 3586-3594. Silverstein, A. B. (1985). "Two- and four-subtest short forms of the WAIS-R: a closer look at validity and reliability." Journal of Clinical Psychology 41(1): 95-97. Tiihonen, J., J. Haukka, et al. (2005). "Premorbid intellectual functioning in bipolar disorder and schizophrenia: results from a cohort study of male conscripts." American Journal of Psychiatry 162(10): 1904-1910. van Haren, N. E., H. G. Schnack, et al. (2011). "Changes in cortical thickness during the course of illness in schizophrenia." Archives of General Psychiatry 68(9): 871-880. Voets, N. L., M. G. Hough, et al. (2008). "Evidence for abnormalities of cortical development in adolescent-onset schizophrenia." Neuroimage 43(4): 665-675. White, T., N. C. Andreasen, et al. (2003). "Gyrification abnormalities in childhood- and adolescent-onset schizophrenia." Biological Psychiatry 54(4): 418-426. Wilke, M., R. A. Kowatch, et al. (2004). "Voxel-based morphometry in adolescents with bipolar disorder: first results." Psychiatry Research 131(1): 57-69. Winkler, A. M., M. R. Sabuncu, et al. (2012). "Measuring and comparing brain cortical surface area and other areal quantities." Neuroimage 61(4): 1428-1443. Woodberry, K. A., A. J. Giuliano, et al. (2008). "Premorbid IQ in schizophrenia: a meta- analytic review." American Journal of Psychiatry 165(5): 579-587. Young, R. C., J. T. Biggs, et al. (1978). "A rating scale for mania: reliability, validity and sensitivity." British Journal of Psychiatry 133: 429-435. Zilles, K., N. Palomero-Gallagher, et al. (2013). "Development of cortical folding during evolution and ontogeny." Trends in Neurosciences 36(5): 275-284.

102 General discussion and future work

Chapter 6 Conclusions and future work

6.1. Original contributions and general discussion of this thesis

A number of original contributions have been produced as part of this thesis. Some of them constitute new techniques while some others involve combining different pre- existing morphomertric methods in a new fashion. For example, Chapter 3 describes the application of a 3D shape analysis technique, based on spherical harmonic functions, to analyze thalamic shape. This technique employs a fine-scale shape representation of the thalamic surface in order to precisely locate local morphological differences between patients with early onset of psychosis and healthy controls. The automated assessment of total thalamus volume was performed twice, using two different image processing software suits in order to confirm that volume findings were independent of the preprocessing method. In addition, a method for finding the optimal parameters for the spherical harmonics decomposition of the thalamic surfaces was developed. This method consisted of creating thalamic tessellations with different spatial resolutions and degrees of spherical harmonics decomposition and compares thereafter the point-wise distance between the original thalamic surface and each parametric representation. The combination of parameters where the average distance between the SPHARM-based and the original surfaces was lower was chosen as the optimal and was used in the study. This work has underlined the importance of shape analysis technique to study, in vivo, morphology of subcortical structures. Although voxel based morphometry can also be used for localizing regional thalamic differences, shape analysis techniques have two main advantages. First, the rigid alignment maintains thalamic surfaces in individual space, thereby avoiding the spatial normalization and modulation steps needed in VBM. Second, interpretation of results is straightforward since the employed metric, point- wise distance (in millimeters) from individual thalamic surfaces to the average surface, has a clear physical meaning. The techniques in Chapter 4 investigated longitudinal changes over time in cortical surface area, thickness, gyrification index, hull surface area, gyral white matter thickness, sulcal depth, length, and width in order to provide a more complete

103 Chapter 6 characterization of cortical development during adolescence and to examine whether the patterns of cortical development were similar in all brain lobes. This study proposed a new and detailed set of sulcal and gyral measurements in addition to traditional surface- based measurements such as thickness and surface area. The obtained results replicated previous reports showing widespread cortical thinning during adolescence. This study was also the first one to show loss of sulcal surface area in males and females during adolescence and demonstrated how this loss was related to changes in sulcal depth, length and width. These results were possible due to the development of tools that enabled combining different image processing packages. Combining different software packages is a challenge mainly due to the fact that packages use different image formats which store image characteristics such as image orientation in different ways. The tools specially designed as part of this thesis facilitated the use of cortical surfaces and tissue segmentations created with FreeSurfer software and port these into BrainVISA sofware for assessment sulcal morphology. Finally, in Chapter 5 some of the cortical biomarkers and comparison strategies from Chapter 4 were applied to compare the cortical morphology of adolescents with first-episode early-onset psychosis and healthy subjects. An innovative aspect of the study is the assessment of the association between sulcal width and the gyrification index. Various new hypotheses regarding the brain dynamics underlying early-onset psychosis can be generated by combining the results presented in the chapters 3, 4, and 5. In adolescents with early-onset psychosis reduced thalamic volume, particularly around the anterior mediodorsal area (discussion section in Chapter 3) could be related to the reduced prefrontal gyrification (discussion section in Chapter 5). The well described anatomical and functional connectivity between the thalamic and prefrontal subregions suggests that a close relationship between the findings presented in chapters 3 and 5 can exist. Both these regions are described as key regions implicated in disease expression, particularly the cognitive impairment from which these patients suffer (Selemon et al. 2005). Speculatively, the simultaneous decrease of thalamic surface area and prefrontal gyrification in patients could be a consequence of an impairment of prefrontothalamic white matter connections (Pakkenberg 1992; Young et al. 2000; Danos et al. 2005). In this context, our findings are an incentive for future studies using

104 General discussion and future work multi-modal imaging to further assess the relationship between thalamic and prefrontal surface shape and its relationship with measures of connectivity in patients with EOP. The decrease in prefrontal cortical thickness and the strong negative correlation between sulcal width and gyrification index in prefrontal regions (Chapter 5) linked to the key role of sulcal width in the cortical flattening process during adolescence (Chapter 4) extends our current knowledge of macroscopic brain changes associated with EOP and their complex interrelationship. Our results strongly suggest that the assessment of multiple morphological parameters provides a more complete picture of the various disease-related macroscopic changes that take place in EOP compared to assessment of only one or a few parameters. These results argue in favor of future clinical neuroscience studies assessing a larger set of morphological parameters. We believe that the assessment and combination of multiple morphological parameters will add to the discovery of better neuroimaging biomarkers, improved patient stratification and treatment. Neuroimaging is one of the few techniques that allow for studying the brain in vivo but this by itself is not sufficient for successful integration in clinical practice. Neuroimaging can only make the transition when it starts to produce meaningful results in aspects such as mentioned above. In this respect, the application of machine learning algorithms using MRI-derived features is gaining interest. In clinical neuroscience, machine learning algorithms are used to predict e.g. diagnosis or age using neuroimaging features. Nevertheless successful prediction rates are still too low to be clinically useful. One of the reasons may be that the majority of machine learning studies use only one or a few morphological features which may be suboptimal. Our pipeline allows for the automatic computation of a large set of morphological features and as such, the integration of our pipeline in large scale machine learning studies may be of interest to the field. Although only results for healthy and early-onset of psychosis adolescents were shown, it is important to highlight that every processing strategy developed in this dissertation can be used in the study of other pathologies such as Alzheimer‟s disease and typical senescence. The application in Alzheimer‟s disease is of particular interest since morphological changes have already been fairly successful in separating those who convert to Alzheimer‟s from those who do not. That is, morphological changes seem to be a core marker of the disease. We have shown that sulcal width may be a marker for cortical atrophy and as such a surrogate for the widely used cortical thickness measurement. The advent of sulcal measurements over cortical thickness is

105 Chapter 6 that these do not depend on the grey-white contrast. This is important because the grey- white contrast is affected by aging itself and therefore any measurement derived from it may be inherently biased. The biomarkers developed and implemented in this thesis may therefore have high applicability of classification studies in Alzheimer‟s disease. All the tools developed and implemented in this thesis have been incorporated into the image processing platform of CIBERSAM (Centro de Investigación Biomédica en Red de Salud Mental, see Appendix 6.A). This platform allows members of CIBERSAM to provide their data for use with our pipeline. We have already successfully processed images of more than 3000 individuals.

6.2. Conclusions

During this dissertation, several image processing strategies were developed to combine different new morphometric techniques and biomarkers in order to study and characterize human brain anatomy. Summarizing, the main conclusions arising from this thesis are: 1. A shape analysis approach is a useful technique for quantifying global and regional thalamic volume differences between patients with psychosis and healthy controls. a. Male adolescents with early-onset first-episode psychosis showed bilateral global thalamus volumetric deficits. b. Patients demonstrated statistically significant right-sided regional thalamic volume differences in areas corresponding to the anterior mediodorsal and pulvinar nuclei when compared with the control group. 2. Different but complementary cortical measurements were combined under a new approach to obtain a more complete characterization about typical development of the human cortex during adolescence. a. A global sulcal widening as well as a decrease in sulcal depth in the frontal and occipital cortex take place over time during adolescence. b. Gyral white matter thickness increases over time in all lobes suggesting that previously-described lobar increases in white matter volume also take place in the gyri adjacent to the cortex. c. The relationships among (a) and (b) were compared directly, providing new insights into the dynamics of macrostructural change during adolescence.

106 General discussion and future work

3. Different lobar biomarkers were simultaneously employed to assess the mechanisms underlying the decrease in cortical thickness and gyrification observed in patients with a psychotic disorder during adolescence. a. A decreased frontal cortical thickness and increased sulcal width in adolescents with psychosis when compared to controls was found. b. Increased sulcal width is associated with a lower frontal gyrification index in the combined patient group. c. In adolescents with follow-up diagnoses of schizophrenia, sulcal width of the temporal, parietal and occipital cortices were increased at the initial stages of the disease suggesting that at an early phase in the disorder there is involvement of multiple distributed brain regions, while for bipolar disorder with psychotic symptoms cortical involvement is more confined to the prefrontal regions. 4. Processing strategies and biomarkers can be easily extrapolated to the study of other brain pathologies.

6.3. Future research lines

Future work is needed in order to address some of the limitations mentioned throughout this thesis. 1. Diffusion scalar maps (i.e., fractional anisotropy or mean diffusivity maps) and fiber tracking, computed via diffusion weighted imaging, could clarify the relationship between prefrontal and thalamic impairments. Further experiments will be carried out to clarify the inferred relationship between the thalamic volume and gyrification index in adolescents with early-onset first-episode psychosis and compare it to the relationship observed in healthy subjects. 2. A replication of the studies, presented at chapters 4 and 5, in larger and homogenous samples will be performed. 3. Diffusion weighted imaging will be used to investigate if the changes in gyral white matter thickness, described in chapter 4, are related to changes in the diffusion properties of gyral white matter tracts (extrinsic white matter tracts near the cortex). 4. The methodology presented in Chapter 4 and 5 will be applied to look for possible differences at higher spatial scales instead of cortical lobes.

107 Chapter 6

Related appendices

Appendix 6.A. Image processing platform

This appendix provides a brief description of the hardware and software used during the development of this thesis.

Hardware

The overall scheme of the computational facilities dedicated to image processing, which are located at Laboratorio de Imagen Médica (LIM) in Hospital General Universitario Gregorio Marañón, is depicted in the following figure (Figure 6.A.1).

Figure 6.A.1: Schematic representation of hardware organization.

The technical characteristics of the personal computer as well as the processing mainstreams are:

1. Personal Computer Model: Dell Precision T3400 with 1 TB of storage capacity. CPU: Intel(R) Core(TM)2 Quad CPU Q9550 @ 2.83GHz. 64 bits. 1333MHz Memory: Kingston 8GB RAM. DIMM DDR2 Synchronous 800 MHz (1.2 ns) Video Card: nVidia G86 [Quadro NVS 290]. PCI-e.256 MB. 64 bits. 33MHz. Storage: 1 TB.

108 General discussion and future work

2. Two processing mainstreams. a. Model: Dell Power Edge R810. CPU: 4 x Intel Xeon E7-4820 (2 GHz x 8 cores). Memory: 192 GB RAM. Storage: 6 TB Raid 5 (3 TB available). b. Model: Dell Power Edge R815. CPU: 2 x AMD Opteron 6238 (2,6 GHz x 48). Memory: 256 GB de RAM. Storage: 6 TB Raid 5 (3 TB available). 3. 60 Virtual Machines with 1 core (30 x 2GHz & 42 x 2‟6 GHZ), 6 GB RAM and 50 GB of storage capacity.

Software

An overview of the CIBERSAM image processing platform which includes main structural and functional images processing stages as well as the image storage platform is shown in Figure 6.A.2.

Figure 6.A.2: Schematic representation of CIBERSAM image processing platform.

The image storage platform was designed as an open and modular architecture that allows for defining multiple configurations about receiving, processing and storage medical images in DICOM format. It includes different configurable stages which

109 Chapter 6 interconnect in different ways, depending on the specific requirements of each user or project. Some of these stages are DICOM anonimization, DICOM storage and organization and DICOM to Nifti-1 file conversion. On the other hand the image processing platform is responsible for every image processing workflow applied to each individual image. This global processing workflow can be divided in two main processing streams, one for structural images (T1, T2 and diffusion weighted (DW) images) and another for functional images. The structural processing pipelines are created using some of the most used software packages (FreeSurfer, FSL TrackVis, Diffusion Toolkit, BrainVISA, ANTs and SPM). The components of all these software packages can be combined in various ways to create new processing workflows apart from the standard ones. For each subject a configuration file that contains all parameters required by the processing is created. The configuration file specifies and controls the processing workflow and also fulfills the role of "log" file. The configuration file also allows both results traceability and results reproducibility. Every processing workflow is oriented to individual subjects, thus the “parallelization” is achieved by sending each processing to a different virtual machine. The local computer and the remote machines of the cluster are connected using Secure Shell (SSH) protocol. Each user of the platform has access to the cluster via its own personal computer using a Graphical User Interface (GUI). Using this interface, the user can remotely select the processes to run and the VMs needed (and available). The data required for the procedure is located on a common server that has direct connection with the cluster in order to enhance the efficiency of the data transfer. The users can access this server from their local computers. Once the procedure has finished, the results are stored back on the server and the unnecessary data is deleted. Currently the platform has different categories of users, which can be broadly classified according to their knowledge in neuroimaging and their programming skills. Those with a good background in neuroimaging and able to write sophisticated algorithms (mostly engineers) have access to the platform to create workflows for data processing and to develop new methods of imaging analysis. The users with good knowledge in neuroimaging but low programming skills (typically neuroscientist) usually access to the platform to test their data for a specific hypothesis and their programming skills allows them to adjust previously written scripts such that these fit their data and analysis/hypothesis. Finally, another important group of users are

110 General discussion and future work clinicians (clinical psychologist or psychiatrist) or neuroscience students with no knowledge about programming and different degrees of knowledge about neuroimaging. As users of the platform, this last group usually works together with a member of any of the two groups mentioned above and use a Graphical User Interface with a very detailed description of the process or both.

Software summary

These software packages have been used to develop image processing workflows:

Operating Systems: Ubuntu 11.04 (Natty Narwhal) VMWare Vsphere ESXi 5.5.0. Debian 8 Development Tools: MATLAB 2011 Matworks Bash Python 2.6.6 Image Processing Packages: FSL (v5.0) : http://fsl.fmrib.ox.ac.uk Freesurfer (v5.1 & v5.3): http://surfer.nmr.mgh.harvard.edu/fswiki BrainVisa (v4.4): http://brainvisa.info TrackVis (v0.5.2) and Diffusion Toolkit (0.6.2): http://trackvis.org CAMINO : http://cmic.cs.ucl.ac.uk/camino/ ANTS : http://stnava.github.io/ANTs/ SPM8 (2008b) : http://www.fil.ion.ucl.ac.uk/spm/ Connectome mapper (v2.0) : http://www.cmtk.org/mapper/ Statistics Software package: SPSS (v.13.0). Image Visualization Tools: FSLview 4.0.1 Itksnap 2.4.0 Mricron 2011

111 Chapter 6

Related references

Danos, P., A. Schmidt, et al. (2005). "Volume and neuron number of the mediodorsal thalamic nucleus in schizophrenia: a replication study." Psychiatry Research 140(3): 281-289. Pakkenberg, B. (1992). "The volume of the mediodorsal thalamic nucleus in treated and untreated schizophrenics." Schizophrenia Research 7(2): 95-100. Selemon, L. D., L. Wang, et al. (2005). "Direct and indirect effects of fetal irradiation on cortical gray and white matter volume in the macaque." Biological Psychiatry 57(1): 83-90. Young, K. A., K. F. Manaye, et al. (2000). "Reduced number of mediodorsal and anterior thalamic neurons in schizophrenia." Biological Psychiatry 47(11): 944-953.

112 List of publications

List of publications

Publications in peer- reviewed journals related with this thesis Joost Janssen*, Yasser Alemán-Gómez*, Hugo Schnack, Evan Balaban, Laura Pina-Camacho, Fidel Alfaro-Almagro, Josefina Castro-Fornieles, Soraya Otero, Inmaculada Baeza, Dolores Moreno, Nuria Bargalló, Mara Parellada, Celso Arango, Manuel Desco: Cortical morphology of adolescents with bipolar disorder and with schizophrenia. Schizophrenia Research 07/2014; 158(1-3). * Means equal contribution. Yasser Alemán-Gómez*, Joost Janssen*, Hugo Schnack, Evan Balaban, Laura Pina-Camacho, Fidel Alfaro-Almagro, Josefina Castro-Fornieles, Soraya Otero, Immaculada Baeza, Dolores Moreno, Nuria Bargalló, Mara Parellada, Celso Arango, Manuel Desco: The Human Cerebral Cortex Flattens during Adolescence. Journal of Neuroscience. 09/2013; 33(38):15004-15010. * Means equal contribution. Joost Janssen, Yasser Alemán-Gómez, Santiago Reig, Hugo G Schnack, Mara Parellada, Montserrat Graell, Carmen Moreno, Dolores Moreno, J M Mateos-Pérez, J M Udias, Celso Arango, Manuel Desco: Regional specificity of thalamic volume deficits in male adolescents with early-onset psychosis. The British journal of psychiatry. 11/2011; 200(1):30-6. Joost Janssen, Santiago Reig, Yasser Alemán-Gómez, Hugo Schnack, J M Udias, Mara Parellada, Montserrat Graell, Dolores Moreno, Arantzazu Zabala, Evan Balaban, Manuel Desco, Celso Arango: Gyral and Sulcal Cortical Thinning in Adolescents with First Episode Early-Onset Psychosis. Biological psychiatry 08/2009; 66(11):1047-54.

Presentations in international conferences related with this thesis Yasser Alemán-Gómez, J Janssen, H Schnack, E Balaban, L Pina-Camacho, F Alfaro- Almagro, J Castro-Fornieles, S Otero, I Baeza, D Moreno, N Bargalló, M Parellada, C Arango, M Desco. 2013. The Human Cortex Flattens during Adolescence. Accepted at the 19th Annual Meeting of the Organization for Human Brain Mapping, June 16-20, 2013, Seattle, WA, USA. Available on CD-Room in NeuroImage. Janssen J., Alemán-Gómez Yasser., Reig S., Schnack H., Parellada M., Graell M., Moreno C., Moreno D., Mateos-Pérez J.M., Udías J.M., Desco M., Arango C.,2010. Thalamus volume and shape in male adolescents with early-onset first-episode psychosis, Schizophr Res. pp. 203. Janssen J., Alemán-Gómez Yasser., Reig S., Schnack H., Parellada M., Graell M., Moreno C., Moreno D., Mateos-Pérez J.M., Udías J.M., Desco M., Arango C.,2010. Thalamus volume and shape in male adolescents with early-onset first-episode psychosis, 2nd Conference of the Schizophrenia International Research Society (SIRS), Florence, Italy. Janssen J., Reig S., Alemán-Gómez Yasser, Navas J., Moreno D., Parellada M., Desco M., Arango C.,2010. Morphometry of the superior frontal cortex in adolescent-onset psychosis, 16th Annual Meeting of the Organization for Human Brain Mapping (HBM 2010), Barcelona.

Other publications in peer-reviewed journals Erick J. Canales-Rodríguez, Alessandro Daducci, Stamatios N. Sotiropoulos, Emmanuel Caruyer, Santiago Aja-Fernández, Joaquim Radua, Jesús M. Yurramendi Mendizabal, Yasser Iturria-Medina, Lester Melie-García, Yasser Alemán-Gómez, Jean-Philippe Thiran, Salvador Sarró, Edith Pomarol-Clotet, and Raymond Salvador. Spherical deconvolution of multichannel diffusion MRI data with non-Gaussian noise models and spatial regularization. PLoS ONE. 9/2015. (Accepted);

113 List of publications

Cristina Chavarrías, Verónica García-Vázquez, Yasser Alemán-Gómez, Paula Montesinos, Javier Pascau, Manuel Desco: fMRat: an extension of SPM for a fully automatic analysis of rodent brain functional magnetic resonance series. Medical & Biological Engineering 08/2015; Alejandro Pérez Fernández, Lorna García-Pentón, Erick J. Canales-Rodríguez, Garikoitz Lerma-Usabiaga, Yasser Iturria-Medina, Francisco Javier Román, Doug Davidson, Yasser Alemán-Gómez, Joana Acha, Manuel Carreiras: Brain morphometry of Dravet Syndrome. Epilepsy Research 10/2014; 108:1326-1334. Francisco J Navas-Sánchez, Yasser Alemán-Gómez, Javier Sánchez-Gonzalez, Juan A Guzmán-De-Villoria, Carolina Franco, Olalla Robles, Celso Arango, Manuel Desco: White matter microstructure correlates of mathematical giftedness and intelligence quotient. Human Brain Mapping 06/2014; 35(6). María Lacalle-Aurioles, José M Mateos-Pérez, Juan A Guzmán-De-Villoria, Javier Olazarán, Isabel Cruz-Orduña, Yasser Alemán-Gómez, María-Elena Martino, Manuel Desco: Cerebral blood flow is an earlier indicator of perfusion abnormalities than cerebral blood volume in Alzheimer's disease. Journal of cerebral blood flow and metabolism. 01/2014; 34(4). María Lacalle-Aurioles, Yasser Alemán-Gómez, Juan Adán Guzmán-De-Villoria, Isabel Cruz- Orduña, Javier Olazarán, José María Mateos-Pérez, María Elena Martino, Manuel Desco: Is the Cerebellum the Optimal Reference Region for Intensity Normalization of Perfusion MR Studies in Early Alzheimer's Disease?. PLoS ONE 12/2013; 8(12):e81548. Erick Jorge Canales-Rodríguez, Joaquim Radua, Edith Pomarol-Clotet, Salvador Sarró, Yasser Alemán-Gómez, Yasser Iturria-Medina, Raymond Salvador: Statistical Analysis of Brain Tissue Images in the Wavelet Domain: Wavelet-Based Morphometry.. NeuroImage 02/2013; 72. Gretel Sanabria-Diaz, Lester Melie-García, Yasser Iturria-Medina, Yasser Alemán-Gómez, Gertrudis Hernández-González, Lourdes Valdés-Urrutia, Lídice Galán, Pedro Valdés- Sosa: Surface area and cortical thickness descriptors reveal different attributes of the structural human brain networks. NeuroImage 05/2010; 50(4):1497-510. Erick Jorge Canales-Rodríguez, Yasser Iturria-Medina, Yasser Alemán-Gómez, Lester Melie- García: Deconvolution in diffusion spectrum imaging. NeuroImage 12/2009; 50(1):136- 49. Pedro A Valdes-Sosa, Jose Miguel Sanchez-Bornot, Roberto Carlos Sotero, Yasser Iturria- Medina, Yasser Aleman-Gomez, Jorge Bosch-Bayard, Felix Carbonell, Tohru Ozaki: Model driven EEG/fMRI fusion of brain oscillations. Human Brain Mapping 09/2009; 30(9):2701-21. Pedro A Valdés-Hernández, Nicolás von Ellenrieder, Alejandro Ojeda-Gonzalez, Silvia Kochen, Yasser Alemán-Gómez, Carlos Muravchik, Pedro A Valdés-Sosa: Approximate average head models for EEG source imaging. Journal of neuroscience methods 09/2009; 185(1):125-32. Erick Jorge Canales-Rodríguez, Lester Melie-García, Yasser Iturria-Medina, Eduardo Martínez- Montes, Yasser Alemán-Gómez, Ching-Po Lin: Inferring multiple maxima in intravoxel white matter fiber distribution. Magnetic Resonance in Medicine 09/2008; 60(3):616-30. Yasser Iturria-Medina, Roberto C Sotero, Erick J Canales-Rodríguez, Yasser Alemán-Gómez, Lester Melie-García: Studying the human brain anatomical network via diffusion- weighted MRI and Graph Theory. NeuroImage 05/2008; 40(3):1064-76.

114 List of publications

Lester Melie-García, Erick J Canales-Rodríguez, Yasser Alemán-Gómez, Ching-Po Lin, Yasser Iturria-Medina, Pedro A Valdés-Hernández: A Bayesian framework to identify principal intravoxel diffusion profiles based on diffusion-weighted MR imaging. NeuroImage 05/2008; 42(2):750-70. Y Iturria-Medina, E J Canales-Rodríguez, L Melie-García, P.A. Valdés-Hernández, E Martínez- Montes, Yasser Alemán-Gómez, J M Sánchez-Bornot: Characterizing brain anatomical connections using diffusion weighted MRI and graph theory. NeuroImage 08/2007; 36(3):645-60.

Selected publications in international conferences Erick Jorge Canales-Rodríguez, Lester Melie-García, Yasser Iturria-Medina, Yasser Alemán- Gómez: Evaluation of the Deconvolved Diffusion Spectrum Imaging Technique. IEEE International Symposium on Biomedical Imaging (ISBI 2012), workshop on high angular resolution diffusion MRI reconstruction techniques; 05/2012 Erick Jorge Canales-Rodríguez, Lester Melie-García, Yasser Iturria-Medina, Yasser Alemán- Gómez: Multi-Tensor Fitting Guided by ODF Estimation. IEEE International Symposium on Biomedical Imaging (ISBI 2012), workshop on high angular resolution diffusion MRI reconstruction techniques; 05/2012 Alemán-Gómez Yasser, Santiago Reig, Joost Janssen, Lester Melie-García, Manuel Desco- Menéndez. 2010. Evaluation of ANIMAL, IBASPM, FIRST and FreeSurfer for subcortical parcellation. Accepted at the 16th Annual Meeting of the Organization for Human Brain Mapping, June 6-10, 2010, Barcelona, Spain. Available on CD-Room in NeuroImage.

115